X-Git-Url: https://git.octo.it/?a=blobdiff_plain;f=src%2Fnetapp.c;h=ee8bfcf0383a154684feb754ab63a43ddc22b0c2;hb=96d86b57692122d76e3687a3464971f1d377eec7;hp=d5eeb6c1a648b419723d3a843622b9de240a6948;hpb=235e731905fc81a379ba7ba0717dd96a69ac2737;p=collectd.git diff --git a/src/netapp.c b/src/netapp.c index d5eeb6c1..ee8bfcf0 100644 --- a/src/netapp.c +++ b/src/netapp.c @@ -29,91 +29,116 @@ #include +#define HAS_ALL_FLAGS(has,needs) (((has) & (needs)) == (needs)) + typedef struct host_config_s host_config_t; typedef void service_handler_t(host_config_t *host, na_elem_t *result, void *data); -#define PERF_SYSTEM_CPU 0x01 -#define PERF_SYSTEM_NET 0x02 -#define PERF_SYSTEM_OPS 0x04 -#define PERF_SYSTEM_DISK 0x08 -#define PERF_SYSTEM_ALL 0x0F - /*! - * \brief Persistent data for system performence counters + * \brief Persistent data for system performance counters */ - +#define CFG_SYSTEM_CPU 0x01 +#define CFG_SYSTEM_NET 0x02 +#define CFG_SYSTEM_OPS 0x04 +#define CFG_SYSTEM_DISK 0x08 +#define CFG_SYSTEM_ALL 0x0F typedef struct { uint32_t flags; - uint64_t last_cpu_busy; - uint64_t last_cpu_total; -} perf_system_data_t; +} cfg_system_t; /*! - * \brief Persistent data for WAFL performence counters. (a.k.a. cache performence) + * \brief Persistent data for WAFL performance counters. (a.k.a. cache performance) + * + * The cache counters use old counter values to calculate a hit ratio for each + * counter. The "data_wafl_t" struct therefore contains old counter values + * along with flags, which are set if the counter is valid. + * + * The function "query_wafl_data" will fill a new structure of this kind with + * new values, then pass both, new and old data, to "submit_wafl_data". That + * function calculates the hit ratios, submits the calculated values and + * updates the old counter values for the next iteration. */ - -#define PERF_WAFL_NAME_CACHE 0x01 -#define PERF_WAFL_DIR_CACHE 0x02 -#define PERF_WAFL_BUF_CACHE 0x04 -#define PERF_WAFL_INODE_CACHE 0x08 -#define PERF_WAFL_ALL 0x0F - +#define CFG_WAFL_NAME_CACHE 0x0001 +#define CFG_WAFL_DIR_CACHE 0x0002 +#define CFG_WAFL_BUF_CACHE 0x0004 +#define CFG_WAFL_INODE_CACHE 0x0008 +#define CFG_WAFL_ALL 0x000F +#define HAVE_WAFL_NAME_CACHE_HIT 0x0100 +#define HAVE_WAFL_NAME_CACHE_MISS 0x0200 +#define HAVE_WAFL_NAME_CACHE (HAVE_WAFL_NAME_CACHE_HIT | HAVE_WAFL_NAME_CACHE_MISS) +#define HAVE_WAFL_FIND_DIR_HIT 0x0400 +#define HAVE_WAFL_FIND_DIR_MISS 0x0800 +#define HAVE_WAFL_FIND_DIR (HAVE_WAFL_FIND_DIR_HIT | HAVE_WAFL_FIND_DIR_MISS) +#define HAVE_WAFL_BUF_HASH_HIT 0x1000 +#define HAVE_WAFL_BUF_HASH_MISS 0x2000 +#define HAVE_WAFL_BUF_HASH (HAVE_WAFL_BUF_HASH_HIT | HAVE_WAFL_BUF_HASH_MISS) +#define HAVE_WAFL_INODE_CACHE_HIT 0x4000 +#define HAVE_WAFL_INODE_CACHE_MISS 0x8000 +#define HAVE_WAFL_INODE_CACHE (HAVE_WAFL_INODE_CACHE_HIT | HAVE_WAFL_INODE_CACHE_MISS) +#define HAVE_WAFL_ALL 0xff00 typedef struct { uint32_t flags; - uint64_t last_name_cache_hit; - uint64_t last_name_cache_miss; - uint64_t last_find_dir_hit; - uint64_t last_find_dir_miss; - uint64_t last_buf_hash_hit; - uint64_t last_buf_hash_miss; - uint64_t last_inode_cache_hit; - uint64_t last_inode_cache_miss; -} perf_wafl_data_t; - -#define PERF_VOLUME_INIT 0x01 -#define PERF_VOLUME_IO 0x02 -#define PERF_VOLUME_OPS 0x03 -#define PERF_VOLUME_LATENCY 0x08 -#define PERF_VOLUME_ALL 0x0F - -typedef struct { - uint32_t flags; -} perf_volume_data_t; - -typedef struct { - uint32_t flags; -} volume_data_t; - -#define PERF_DISK_BUSIEST 0x01 -#define PERF_DISK_ALL 0x01 + time_t timestamp; + uint64_t name_cache_hit; + uint64_t name_cache_miss; + uint64_t find_dir_hit; + uint64_t find_dir_miss; + uint64_t buf_hash_hit; + uint64_t buf_hash_miss; + uint64_t inode_cache_hit; + uint64_t inode_cache_miss; +} data_wafl_t; +/*! + * \brief Persistent data for volume performance data. + * + * The code below uses the difference of the operations and latency counters to + * calculate an average per-operation latency. For this, old counters need to + * be stored in the "data_volume_perf_t" structure. The byte-counters are just + * kept for completeness sake. The "flags" member indicates if each counter is + * valid or not. + * + * The "query_volume_perf_data" function will fill a new struct of this type + * and pass both, old and new data, to "submit_volume_perf_data". In that + * function, the per-operation latency is calculated and dispatched, then the + * old counters are updated. + */ +#define CFG_VOLUME_PERF_INIT 0x0001 +#define CFG_VOLUME_PERF_IO 0x0002 +#define CFG_VOLUME_PERF_OPS 0x0003 +#define CFG_VOLUME_PERF_LATENCY 0x0008 +#define CFG_VOLUME_PERF_ALL 0x000F +#define HAVE_VOLUME_PERF_BYTES_READ 0x0010 +#define HAVE_VOLUME_PERF_BYTES_WRITE 0x0020 +#define HAVE_VOLUME_PERF_OPS_READ 0x0040 +#define HAVE_VOLUME_PERF_OPS_WRITE 0x0080 +#define HAVE_VOLUME_PERF_LATENCY_READ 0x0100 +#define HAVE_VOLUME_PERF_LATENCY_WRITE 0x0200 +#define HAVE_VOLUME_PERF_ALL 0x03F0 typedef struct { uint32_t flags; -} perf_disk_data_t; +} cfg_volume_perf_t; typedef struct { uint32_t flags; - time_t last_timestamp; - uint64_t last_read_latency; - uint64_t last_write_latency; - uint64_t last_read_ops; - uint64_t last_write_ops; -} per_volume_perf_data_t; + time_t timestamp; + uint64_t read_bytes; + uint64_t write_bytes; + uint64_t read_ops; + uint64_t write_ops; + uint64_t read_latency; + uint64_t write_latency; +} data_volume_perf_t; +/*! + * \brief Configuration struct for volume usage data (free / used). + */ #define VOLUME_INIT 0x01 #define VOLUME_DF 0x02 #define VOLUME_SNAP 0x04 - typedef struct { uint32_t flags; -} per_volume_data_t; - -typedef struct { - time_t last_update; - double last_disk_busy_percent; - uint64_t last_disk_busy; - uint64_t last_base_for_disk_busy; -} per_disk_perf_data_t; +} cfg_volume_usage_t; typedef struct service_config_s { na_elem_t *query; @@ -123,32 +148,51 @@ typedef struct service_config_s { int interval; void *data; struct service_config_s *next; -} service_config_t; - +} cfg_service_t; #define SERVICE_INIT {0, 0, 1, 1, 0, 0, 0} +/*! + * \brief Struct representing a volume. + * + * A volume currently has a name and two sets of values: + * + * - Performance data, such as bytes read/written, number of operations + * performed and average time per operation. + * + * - Usage data, i. e. amount of used and free space in the volume. + */ typedef struct volume_s { char *name; - per_volume_perf_data_t perf_data; - per_volume_data_t volume_data; + data_volume_perf_t perf_data; + cfg_volume_usage_t cfg_volume_usage; struct volume_s *next; } volume_t; +#define CFG_DISK_BUSIEST 0x01 +#define CFG_DISK_ALL 0x01 +#define HAVE_DISK_BUSY 0x10 +#define HAVE_DISK_BASE 0x20 +#define HAVE_DISK_ALL 0x30 +typedef struct { + uint32_t flags; +} cfg_disk_t; + /*! - * \brief A disk in the netapp. + * \brief A disk in the NetApp. * - * A disk doesn't have any more information than its name atm. + * A disk doesn't have any more information than its name at the moment. * The name includes the "disk_" prefix. */ - typedef struct disk_s { char *name; - per_disk_perf_data_t perf_data; + uint32_t flags; + time_t timestamp; + uint64_t disk_busy; + uint64_t base_for_disk_busy; + double disk_busy_percent; struct disk_s *next; } disk_t; -#define DISK_INIT {0, {0, 0, 0, 0}, 0} - struct host_config_s { na_server_t *srv; char *name; @@ -158,385 +202,708 @@ struct host_config_s { char *username; char *password; int interval; - service_config_t *services; + cfg_service_t *services; disk_t *disks; volume_t *volumes; struct host_config_s *next; }; - -#define HOST_INIT {0, 0, NA_SERVER_TRANSPORT_HTTPS, 0, 0, 0, 0, 10, 0, 0, 0} +#define HOST_INIT {NULL, NULL, NA_SERVER_TRANSPORT_HTTPS, NULL, 0, NULL, NULL, 10, NULL, NULL, NULL, NULL} static host_config_t *host_config; -static volume_t *get_volume(host_config_t *host, const char *name) { +/* + * Auxiliary functions + * + * Used to look up volumes and disks or to handle flags. + */ +static volume_t *get_volume (host_config_t *host, const char *name, /* {{{ */ + uint32_t vol_usage_flags, uint32_t vol_perf_flags) +{ volume_t *v; + + if (name == NULL) + return (NULL); + /* Make sure the default flags include the init-bit. */ + if (vol_usage_flags != 0) + vol_usage_flags |= VOLUME_INIT; + if (vol_perf_flags != 0) + vol_perf_flags |= CFG_VOLUME_PERF_INIT; + for (v = host->volumes; v; v = v->next) { - if (!strcmp(v->name/* + 4*/, name)) return v; + if (strcmp(v->name, name) != 0) + continue; + + /* Check if the flags have been initialized. */ + if (((v->cfg_volume_usage.flags & VOLUME_INIT) == 0) + && (vol_usage_flags != 0)) + v->cfg_volume_usage.flags = vol_usage_flags; + if (((v->perf_data.flags & CFG_VOLUME_PERF_INIT) == 0) + && (vol_perf_flags != 0)) + v->perf_data.flags = vol_perf_flags; + + return v; } + + DEBUG ("netapp plugin: Allocating new entry for volume %s.", name); v = malloc(sizeof(*v)); + if (v == NULL) + return (NULL); + memset (v, 0, sizeof (*v)); + + v->cfg_volume_usage.flags = vol_usage_flags; + v->perf_data.flags = vol_perf_flags; + v->name = strdup(name); -// v->name = malloc(strlen(name) + 5); -// strcpy(v->name, "vol-"); -// strcpy(v->name + 4, name); - v->perf_data.flags = 0; - v->volume_data.flags = 0; + if (v->name == NULL) { + sfree (v); + return (NULL); + } + v->next = host->volumes; host->volumes = v; + return v; -} +} /* }}} volume_t *get_volume */ + +static disk_t *get_disk(host_config_t *host, const char *name) /* {{{ */ +{ + disk_t *v; -static disk_t *get_disk(host_config_t *host, const char *name) { - disk_t *v, init = DISK_INIT; + if (name == NULL) + return (NULL); for (v = host->disks; v; v = v->next) { - if (!strcmp(v->name/* + 5*/, name)) return v; + if (strcmp(v->name, name) == 0) + return v; } v = malloc(sizeof(*v)); - *v = init; + if (v == NULL) + return (NULL); + memset (v, 0, sizeof (*v)); + v->next = NULL; + v->name = strdup(name); -// v->name = malloc(strlen(name) + 6); -// strcpy(v->name, "disk-"); -// strcpy(v->name + 5, name); + if (v->name == NULL) { + sfree (v); + return (NULL); + } + v->next = host->disks; host->disks = v; + return v; -} +} /* }}} disk_t *get_disk */ -static void collect_perf_wafl_data(host_config_t *host, na_elem_t *out, void *data) { - perf_wafl_data_t *wafl = data; - uint64_t name_cache_hit = 0, name_cache_miss = 0, find_dir_hit = 0, find_dir_miss = 0, buf_hash_hit = 0, buf_hash_miss = 0, inode_cache_hit = 0, inode_cache_miss = 0; - const char *instance, *name; - time_t timestamp; - na_elem_t *counter; - value_t values[2]; +static void set_global_perf_vol_flag(const host_config_t *host, /* {{{ */ + uint32_t flag, _Bool set) +{ + volume_t *v; + + for (v = host->volumes; v; v = v->next) { + if (set) + v->perf_data.flags |= flag; + else /* if (!set) */ + v->perf_data.flags &= ~flag; + } +} /* }}} void set_global_perf_vol_flag */ + +static void set_global_vol_flag(const host_config_t *host, /* {{{ */ + uint32_t flag, _Bool set) { + volume_t *v; + + for (v = host->volumes; v; v = v->next) { + if (set) + v->cfg_volume_usage.flags |= flag; + else /* if (!set) */ + v->cfg_volume_usage.flags &= ~flag; + } +} /* }}} void set_global_vol_flag */ + +/* + * Various submit functions. + * + * They all eventually call "submit_values" which creates a value_list_t and + * dispatches it to the daemon. + */ +static int submit_values (const char *host, /* {{{ */ + const char *plugin_inst, + const char *type, const char *type_inst, + value_t *values, int values_len, + time_t timestamp) +{ value_list_t vl = VALUE_LIST_INIT; + + vl.values = values; + vl.values_len = values_len; + + if (timestamp > 0) + vl.time = timestamp; + + if (host != NULL) + sstrncpy (vl.host, host, sizeof (vl.host)); + else + sstrncpy (vl.host, hostname_g, sizeof (vl.host)); + sstrncpy (vl.plugin, "netapp", sizeof (vl.plugin)); + if (plugin_inst != NULL) + sstrncpy (vl.plugin_instance, plugin_inst, sizeof (vl.plugin_instance)); + sstrncpy (vl.type, type, sizeof (vl.type)); + if (type_inst != NULL) + sstrncpy (vl.type_instance, type_inst, sizeof (vl.type_instance)); + + return (plugin_dispatch_values (&vl)); +} /* }}} int submit_uint64 */ + +static int submit_two_counters (const char *host, const char *plugin_inst, /* {{{ */ + const char *type, const char *type_inst, counter_t val0, counter_t val1, + time_t timestamp) +{ + value_t values[2]; + + values[0].counter = val0; + values[1].counter = val1; + + return (submit_values (host, plugin_inst, type, type_inst, + values, 2, timestamp)); +} /* }}} int submit_two_counters */ + +static int submit_counter (const char *host, const char *plugin_inst, /* {{{ */ + const char *type, const char *type_inst, counter_t counter, time_t timestamp) +{ + value_t v; + + v.counter = counter; + + return (submit_values (host, plugin_inst, type, type_inst, + &v, 1, timestamp)); +} /* }}} int submit_counter */ + +static int submit_two_gauge (const char *host, const char *plugin_inst, /* {{{ */ + const char *type, const char *type_inst, gauge_t val0, gauge_t val1, + time_t timestamp) +{ + value_t values[2]; + + values[0].gauge = val0; + values[1].gauge = val1; + + return (submit_values (host, plugin_inst, type, type_inst, + values, 2, timestamp)); +} /* }}} int submit_two_gauge */ + +static int submit_double (const char *host, const char *plugin_inst, /* {{{ */ + const char *type, const char *type_inst, double d, time_t timestamp) +{ + value_t v; + + v.gauge = (gauge_t) d; + + return (submit_values (host, plugin_inst, type, type_inst, + &v, 1, timestamp)); +} /* }}} int submit_uint64 */ + +/* Calculate hit ratio from old and new counters and submit the resulting + * percentage. Used by "submit_wafl_data". */ +static int submit_cache_ratio (const char *host, /* {{{ */ + const char *plugin_inst, + const char *type_inst, + uint64_t new_hits, + uint64_t new_misses, + uint64_t old_hits, + uint64_t old_misses, + time_t timestamp) +{ + value_t v; + + if ((new_hits >= old_hits) && (new_misses >= old_misses)) { + uint64_t hits; + uint64_t misses; + + hits = new_hits - old_hits; + misses = new_misses - old_misses; + + v.gauge = 100.0 * ((gauge_t) hits) / ((gauge_t) (hits + misses)); + } else { + v.gauge = NAN; + } + + return (submit_values (host, plugin_inst, "cache_ratio", type_inst, + &v, 1, timestamp)); +} /* }}} int submit_cache_ratio */ + +/* Submits all the caches used by WAFL. Uses "submit_cache_ratio". */ +static int submit_wafl_data (const host_config_t *host, const char *instance, /* {{{ */ + data_wafl_t *old_data, const data_wafl_t *new_data) +{ + /* Submit requested counters */ + if (HAS_ALL_FLAGS (old_data->flags, CFG_WAFL_NAME_CACHE | HAVE_WAFL_NAME_CACHE) + && HAS_ALL_FLAGS (new_data->flags, HAVE_WAFL_NAME_CACHE)) + submit_cache_ratio (host->name, instance, "name_cache_hit", + new_data->name_cache_hit, new_data->name_cache_miss, + old_data->name_cache_hit, old_data->name_cache_miss, + new_data->timestamp); + + if (HAS_ALL_FLAGS (old_data->flags, CFG_WAFL_DIR_CACHE | HAVE_WAFL_FIND_DIR) + && HAS_ALL_FLAGS (new_data->flags, HAVE_WAFL_FIND_DIR)) + submit_cache_ratio (host->name, instance, "find_dir_hit", + new_data->find_dir_hit, new_data->find_dir_miss, + old_data->find_dir_hit, old_data->find_dir_miss, + new_data->timestamp); + + if (HAS_ALL_FLAGS (old_data->flags, CFG_WAFL_BUF_CACHE | HAVE_WAFL_BUF_HASH) + && HAS_ALL_FLAGS (new_data->flags, HAVE_WAFL_BUF_HASH)) + submit_cache_ratio (host->name, instance, "buf_hash_hit", + new_data->buf_hash_hit, new_data->buf_hash_miss, + old_data->buf_hash_hit, old_data->buf_hash_miss, + new_data->timestamp); + + if (HAS_ALL_FLAGS (old_data->flags, CFG_WAFL_INODE_CACHE | HAVE_WAFL_INODE_CACHE) + && HAS_ALL_FLAGS (new_data->flags, HAVE_WAFL_INODE_CACHE)) + submit_cache_ratio (host->name, instance, "inode_cache_hit", + new_data->inode_cache_hit, new_data->inode_cache_miss, + old_data->inode_cache_hit, old_data->inode_cache_miss, + new_data->timestamp); + + /* Clear old HAVE_* flags */ + old_data->flags &= ~HAVE_WAFL_ALL; + + /* Copy all counters */ + old_data->timestamp = new_data->timestamp; + old_data->name_cache_hit = new_data->name_cache_hit; + old_data->name_cache_miss = new_data->name_cache_miss; + old_data->find_dir_hit = new_data->find_dir_hit; + old_data->find_dir_miss = new_data->find_dir_miss; + old_data->buf_hash_hit = new_data->buf_hash_hit; + old_data->buf_hash_miss = new_data->buf_hash_miss; + old_data->inode_cache_hit = new_data->inode_cache_hit; + old_data->inode_cache_miss = new_data->inode_cache_miss; + + /* Copy HAVE_* flags */ + old_data->flags |= (new_data->flags & HAVE_WAFL_ALL); + + return (0); +} /* }}} int submit_wafl_data */ + +/* Submits volume performance data to the daemon, taking care to honor and + * update flags appropriately. */ +static int submit_volume_perf_data (const host_config_t *host, /* {{{ */ + volume_t *volume, + const data_volume_perf_t *new_data) +{ + /* Check for and submit disk-octet values */ + if (HAS_ALL_FLAGS (volume->perf_data.flags, CFG_VOLUME_PERF_IO) + && HAS_ALL_FLAGS (new_data->flags, HAVE_VOLUME_PERF_BYTES_READ | HAVE_VOLUME_PERF_BYTES_WRITE)) + { + submit_two_counters (host->name, volume->name, "disk_octets", /* type instance = */ NULL, + (counter_t) new_data->read_bytes, (counter_t) new_data->write_bytes, new_data->timestamp); + } + + /* Check for and submit disk-operations values */ + if (HAS_ALL_FLAGS (volume->perf_data.flags, CFG_VOLUME_PERF_OPS) + && HAS_ALL_FLAGS (new_data->flags, HAVE_VOLUME_PERF_OPS_READ | HAVE_VOLUME_PERF_OPS_WRITE)) + { + submit_two_counters (host->name, volume->name, "disk_ops", /* type instance = */ NULL, + (counter_t) new_data->read_ops, (counter_t) new_data->write_ops, new_data->timestamp); + } + + /* Check for, calculate and submit disk-latency values */ + if (HAS_ALL_FLAGS (volume->perf_data.flags, CFG_VOLUME_PERF_LATENCY + | HAVE_VOLUME_PERF_OPS_READ | HAVE_VOLUME_PERF_OPS_WRITE + | HAVE_VOLUME_PERF_LATENCY_READ | HAVE_VOLUME_PERF_LATENCY_WRITE) + && HAS_ALL_FLAGS (new_data->flags, HAVE_VOLUME_PERF_OPS_READ | HAVE_VOLUME_PERF_OPS_WRITE + | HAVE_VOLUME_PERF_LATENCY_READ | HAVE_VOLUME_PERF_LATENCY_WRITE)) + { + gauge_t latency_per_op_read; + gauge_t latency_per_op_write; + + latency_per_op_read = NAN; + latency_per_op_write = NAN; + + /* Check if a counter wrapped around. */ + if ((new_data->read_ops > volume->perf_data.read_ops) + && (new_data->read_latency > volume->perf_data.read_latency)) + { + uint64_t diff_ops_read; + uint64_t diff_latency_read; + + diff_ops_read = new_data->read_ops - volume->perf_data.read_ops; + diff_latency_read = new_data->read_latency - volume->perf_data.read_latency; + + if (diff_ops_read > 0) + latency_per_op_read = ((gauge_t) diff_latency_read) / ((gauge_t) diff_ops_read); + } + + if ((new_data->write_ops > volume->perf_data.write_ops) + && (new_data->write_latency > volume->perf_data.write_latency)) + { + uint64_t diff_ops_write; + uint64_t diff_latency_write; + + diff_ops_write = new_data->write_ops - volume->perf_data.write_ops; + diff_latency_write = new_data->write_latency - volume->perf_data.write_latency; + + if (diff_ops_write > 0) + latency_per_op_write = ((gauge_t) diff_latency_write) / ((gauge_t) diff_ops_write); + } + + submit_two_gauge (host->name, volume->name, "disk_latency", /* type instance = */ NULL, + latency_per_op_read, latency_per_op_write, new_data->timestamp); + } + + /* Clear all HAVE_* flags. */ + volume->perf_data.flags &= ~HAVE_VOLUME_PERF_ALL; + + /* Copy all counters */ + volume->perf_data.timestamp = new_data->timestamp; + volume->perf_data.read_bytes = new_data->read_bytes; + volume->perf_data.write_bytes = new_data->write_bytes; + volume->perf_data.read_ops = new_data->read_ops; + volume->perf_data.write_ops = new_data->write_ops; + volume->perf_data.read_latency = new_data->read_latency; + volume->perf_data.write_latency = new_data->write_latency; + + /* Copy the HAVE_* flags */ + volume->perf_data.flags |= (new_data->flags & HAVE_VOLUME_PERF_ALL); + + return (0); +} /* }}} int submit_volume_perf_data */ + +/* + * Query functions + * + * These functions are called with appropriate data returned by the libnetapp + * interface which is parsed and submitted with the above functions. + */ +/* Data corresponding to */ +static void query_wafl_data(host_config_t *host, na_elem_t *out, void *data) { /* {{{ */ + data_wafl_t *wafl = data; + data_wafl_t perf_data; + const char *plugin_inst; + na_elem_t *counter; + + memset (&perf_data, 0, sizeof (perf_data)); - timestamp = (time_t) na_child_get_uint64(out, "timestamp", 0); + perf_data.timestamp = (time_t) na_child_get_uint64(out, "timestamp", 0); + out = na_elem_child(na_elem_child(out, "instances"), "instance-data"); - instance = na_child_get_string(out, "name"); + if (out == NULL) + return; + + plugin_inst = na_child_get_string(out, "name"); + if (plugin_inst == NULL) + return; + /* Iterate over all counters */ na_elem_iter_t iter = na_child_iterator(na_elem_child(out, "counters")); for (counter = na_iterator_next(&iter); counter; counter = na_iterator_next(&iter)) { + const char *name; + uint64_t value; + name = na_child_get_string(counter, "name"); + if (name == NULL) + continue; + + value = na_child_get_uint64(counter, "value", UINT64_MAX); + if (value == UINT64_MAX) + continue; + if (!strcmp(name, "name_cache_hit")) { - name_cache_hit = na_child_get_uint64(counter, "value", 0); + perf_data.name_cache_hit = value; + perf_data.flags |= HAVE_WAFL_NAME_CACHE_HIT; } else if (!strcmp(name, "name_cache_miss")) { - name_cache_miss = na_child_get_uint64(counter, "value", 0); + perf_data.name_cache_miss = value; + perf_data.flags |= HAVE_WAFL_NAME_CACHE_MISS; } else if (!strcmp(name, "find_dir_hit")) { - find_dir_hit = na_child_get_uint64(counter, "value", 0); + perf_data.find_dir_hit = value; + perf_data.flags |= HAVE_WAFL_FIND_DIR_HIT; } else if (!strcmp(name, "find_dir_miss")) { - find_dir_miss = na_child_get_uint64(counter, "value", 0); + perf_data.find_dir_miss = value; + perf_data.flags |= HAVE_WAFL_FIND_DIR_MISS; } else if (!strcmp(name, "buf_hash_hit")) { - buf_hash_hit = na_child_get_uint64(counter, "value", 0); + perf_data.buf_hash_hit = value; + perf_data.flags |= HAVE_WAFL_BUF_HASH_HIT; } else if (!strcmp(name, "buf_hash_miss")) { - buf_hash_miss = na_child_get_uint64(counter, "value", 0); + perf_data.buf_hash_miss = value; + perf_data.flags |= HAVE_WAFL_BUF_HASH_MISS; } else if (!strcmp(name, "inode_cache_hit")) { - inode_cache_hit = na_child_get_uint64(counter, "value", 0); + perf_data.inode_cache_hit = value; + perf_data.flags |= HAVE_WAFL_INODE_CACHE_HIT; } else if (!strcmp(name, "inode_cache_miss")) { - inode_cache_miss = na_child_get_uint64(counter, "value", 0); - } - } - if ((wafl->flags & PERF_WAFL_NAME_CACHE) && name_cache_hit && name_cache_miss) { - values[0].gauge = 0; - if (name_cache_miss - wafl->last_name_cache_miss + name_cache_hit - wafl->last_name_cache_hit) values[0].gauge = 100.0 * (name_cache_hit - wafl->last_name_cache_hit) / (name_cache_miss - wafl->last_name_cache_miss + name_cache_hit - wafl->last_name_cache_hit); - vl.values = values; - vl.values_len = 1; - vl.time = timestamp; - vl.interval = interval_g; - sstrncpy(vl.plugin, "netapp", sizeof(vl.plugin)); - sstrncpy(vl.host, host->name, sizeof(vl.host)); - sstrncpy(vl.plugin_instance, instance, sizeof(vl.plugin_instance)); - sstrncpy(vl.type_instance, "name_cache_hit", sizeof(vl.type_instance)); - if (wafl->last_name_cache_hit && wafl->last_name_cache_miss) { - DEBUG("%s/netapp-%s/cache_ratio: %lf", host->name, instance, values[0].gauge); - plugin_dispatch_values ("cache_ratio", &vl); - } - wafl->last_name_cache_hit = name_cache_hit; - wafl->last_name_cache_miss = name_cache_miss; - } - if ((wafl->flags & PERF_WAFL_DIR_CACHE) && find_dir_hit && find_dir_miss) { - values[0].gauge = 0; - if (find_dir_miss - wafl->last_find_dir_miss + find_dir_hit - wafl->last_find_dir_hit) values[0].gauge = 100.0 * (find_dir_hit - wafl->last_find_dir_hit) / (find_dir_miss - wafl->last_find_dir_miss + find_dir_hit - wafl->last_find_dir_hit); - vl.values = values; - vl.values_len = 1; - vl.time = timestamp; - vl.interval = interval_g; - sstrncpy(vl.plugin, "netapp", sizeof(vl.plugin)); - sstrncpy(vl.host, host->name, sizeof(vl.host)); - sstrncpy(vl.plugin_instance, instance, sizeof(vl.plugin_instance)); - sstrncpy(vl.type_instance, "find_dir_hit", sizeof(vl.type_instance)); - if (wafl->last_find_dir_hit && wafl->last_find_dir_miss) { - DEBUG("%s/netapp-%s/cache_ratio: %lf", host->name, instance, values[0].gauge); - plugin_dispatch_values ("cache_ratio", &vl); - } - wafl->last_find_dir_hit = find_dir_hit; - wafl->last_find_dir_miss = find_dir_miss; - } - if ((wafl->flags & PERF_WAFL_BUF_CACHE) && buf_hash_hit && buf_hash_miss) { - values[0].gauge = 0; - if (buf_hash_miss - wafl->last_buf_hash_miss + buf_hash_hit - wafl->last_buf_hash_hit) values[0].gauge = 100.0 * (buf_hash_hit - wafl->last_buf_hash_hit) / (buf_hash_miss - wafl->last_buf_hash_miss + buf_hash_hit - wafl->last_buf_hash_hit); - vl.values = values; - vl.values_len = 1; - vl.time = timestamp; - vl.interval = interval_g; - sstrncpy(vl.plugin, "netapp", sizeof(vl.plugin)); - sstrncpy(vl.host, host->name, sizeof(vl.host)); - sstrncpy(vl.plugin_instance, instance, sizeof(vl.plugin_instance)); - sstrncpy(vl.type_instance, "buf_hash_hit", sizeof(vl.type_instance)); - if (wafl->last_buf_hash_hit && wafl->last_buf_hash_miss) { - DEBUG("%s/netapp-%s/cache_ratio: %lf", host->name, instance, values[0].gauge); - plugin_dispatch_values ("cache_ratio", &vl); - } - wafl->last_buf_hash_hit = buf_hash_hit; - wafl->last_buf_hash_miss = buf_hash_miss; - } - if ((wafl->flags & PERF_WAFL_INODE_CACHE) && inode_cache_hit && inode_cache_miss) { - values[0].gauge = 0; - if (inode_cache_miss - wafl->last_inode_cache_miss + inode_cache_hit - wafl->last_inode_cache_hit) values[0].gauge = 100.0 * (inode_cache_hit - wafl->last_inode_cache_hit) / (inode_cache_miss - wafl->last_inode_cache_miss + inode_cache_hit - wafl->last_inode_cache_hit); - vl.values = values; - vl.values_len = 1; - vl.time = timestamp; - vl.interval = interval_g; - sstrncpy(vl.plugin, "netapp", sizeof(vl.plugin)); - sstrncpy(vl.host, host->name, sizeof(vl.host)); - sstrncpy(vl.plugin_instance, instance, sizeof(vl.plugin_instance)); - sstrncpy(vl.type_instance, "inode_cache_hit", sizeof(vl.type_instance)); - if (wafl->last_inode_cache_hit && wafl->last_inode_cache_miss) { - DEBUG("%s/netapp-%s/cache_ratio: %lf", host->name, instance, values[0].gauge); - plugin_dispatch_values ("cache_ratio", &vl); + perf_data.inode_cache_miss = value; + perf_data.flags |= HAVE_WAFL_INODE_CACHE_MISS; + } else { + DEBUG("netapp plugin: query_wafl_data: Found unexpected child: %s", + name); } - wafl->last_inode_cache_hit = inode_cache_hit; - wafl->last_inode_cache_miss = inode_cache_miss; } -} -static void collect_perf_disk_data(host_config_t *host, na_elem_t *out, void *data) { - perf_disk_data_t *perf = data; - const char *name; + submit_wafl_data (host, plugin_inst, wafl, &perf_data); +} /* }}} void query_wafl_data */ + +/* Data corresponding to */ +static void query_submit_disk_data(host_config_t *host, na_elem_t *out, void *data) { /* {{{ */ + cfg_disk_t *cfg_disk = data; time_t timestamp; na_elem_t *counter, *inst; - disk_t *disk, *worst_disk = 0; - value_t values[2]; - value_list_t vl = VALUE_LIST_INIT; + disk_t *worst_disk = 0; timestamp = (time_t) na_child_get_uint64(out, "timestamp", 0); out = na_elem_child(out, "instances"); + + /* Iterate over all children */ na_elem_iter_t inst_iter = na_child_iterator(out); for (inst = na_iterator_next(&inst_iter); inst; inst = na_iterator_next(&inst_iter)) { - uint64_t disk_busy = 0, base_for_disk_busy = 0; + disk_t *old_data; + disk_t new_data; + + memset (&new_data, 0, sizeof (new_data)); + new_data.timestamp = timestamp; + new_data.disk_busy_percent = NAN; - disk = get_disk(host, na_child_get_string(inst, "name")); + old_data = get_disk(host, na_child_get_string(inst, "name")); + if (old_data == NULL) + continue; + + /* Look for the "disk_busy" and "base_for_disk_busy" counters */ na_elem_iter_t count_iter = na_child_iterator(na_elem_child(inst, "counters")); for (counter = na_iterator_next(&count_iter); counter; counter = na_iterator_next(&count_iter)) { + const char *name; + uint64_t value; + name = na_child_get_string(counter, "name"); - if (!strcmp(name, "disk_busy")) { - disk_busy = na_child_get_uint64(counter, "value", 0); - } else if (!strcmp(name, "base_for_disk_busy")) { - base_for_disk_busy = na_child_get_uint64(counter, "value", 0); + if (name == NULL) + continue; + + value = na_child_get_uint64(counter, "value", UINT64_MAX); + if (value == UINT64_MAX) + continue; + + if (strcmp(name, "disk_busy") == 0) + { + new_data.disk_busy = value; + new_data.flags |= HAVE_DISK_BUSY; + } + else if (strcmp(name, "base_for_disk_busy") == 0) + { + new_data.base_for_disk_busy = value; + new_data.flags |= HAVE_DISK_BASE; } } - if (disk_busy && base_for_disk_busy) { - disk->perf_data.last_update = timestamp; - disk->perf_data.last_disk_busy_percent = 0; - if (base_for_disk_busy - disk->perf_data.last_base_for_disk_busy) disk->perf_data.last_disk_busy_percent = 100.0 * (disk_busy - disk->perf_data.last_disk_busy) / (base_for_disk_busy - disk->perf_data.last_base_for_disk_busy); - if (disk->perf_data.last_disk_busy && disk->perf_data.last_base_for_disk_busy && (!worst_disk || worst_disk->perf_data.last_disk_busy_percent < disk->perf_data.last_disk_busy_percent)) worst_disk = disk; - disk->perf_data.last_disk_busy = disk_busy; - disk->perf_data.last_base_for_disk_busy = base_for_disk_busy; + + /* If all required counters are available and did not just wrap around, + * calculate the busy percentage. Otherwise, the value is initialized to + * NAN at the top of the for-loop. */ + if (HAS_ALL_FLAGS (old_data->flags, HAVE_DISK_BUSY | HAVE_DISK_BASE) + && HAS_ALL_FLAGS (new_data.flags, HAVE_DISK_BUSY | HAVE_DISK_BASE) + && (new_data.disk_busy >= old_data->disk_busy) + && (new_data.base_for_disk_busy > old_data->base_for_disk_busy)) + { + uint64_t busy_diff; + uint64_t base_diff; + + busy_diff = new_data.disk_busy - old_data->disk_busy; + base_diff = new_data.base_for_disk_busy - old_data->base_for_disk_busy; + + new_data.disk_busy_percent = 100.0 + * ((gauge_t) busy_diff) / ((gauge_t) base_diff); } - } - if ((perf->flags & PERF_DISK_BUSIEST) && worst_disk) { - values[0].gauge = worst_disk->perf_data.last_disk_busy_percent; - vl.values = values; - vl.values_len = 1; - vl.time = timestamp; - vl.interval = interval_g; - sstrncpy(vl.plugin, "netapp", sizeof(vl.plugin)); - sstrncpy(vl.host, host->name, sizeof(vl.host)); - sstrncpy(vl.plugin_instance, "system", sizeof(vl.plugin_instance)); - sstrncpy(vl.type_instance, "disk_busy", sizeof(vl.type_instance)); - DEBUG("%s/netapp-system/percent-disk_busy: %lf", host->name, worst_disk->perf_data.last_disk_busy_percent); - plugin_dispatch_values ("percent", &vl); - } -} -static void collect_volume_data(host_config_t *host, na_elem_t *out, void *data) { - na_elem_t *inst, *sis; + /* Clear HAVE_* flags */ + old_data->flags &= ~HAVE_DISK_ALL; + + /* Copy data */ + old_data->timestamp = new_data.timestamp; + old_data->disk_busy = new_data.disk_busy; + old_data->base_for_disk_busy = new_data.base_for_disk_busy; + old_data->disk_busy_percent = new_data.disk_busy_percent; + + /* Copy flags */ + old_data->flags |= (new_data.flags & HAVE_DISK_ALL); + + if ((worst_disk == NULL) + || (worst_disk->disk_busy_percent < old_data->disk_busy_percent)) + worst_disk = old_data; + } /* for (all disks) */ + + if ((cfg_disk->flags & CFG_DISK_BUSIEST) && (worst_disk != NULL)) + submit_double (host->name, "system", "percent", "disk_busy", + worst_disk->disk_busy_percent, timestamp); +} /* }}} void query_submit_disk_data */ + +/* Data corresponding to */ +static void collect_volume_data(host_config_t *host, na_elem_t *out, void *data) { /* {{{ */ + na_elem_t *inst; volume_t *volume; - volume_data_t *volume_data = data; - value_t values[1]; - value_list_t vl = VALUE_LIST_INIT; + cfg_volume_usage_t *cfg_volume_data = data; out = na_elem_child(out, "volumes"); na_elem_iter_t inst_iter = na_child_iterator(out); for (inst = na_iterator_next(&inst_iter); inst; inst = na_iterator_next(&inst_iter)) { - uint64_t size_free = 0, size_used = 0, snap_reserved = 0, sis_saved = 0; - volume = get_volume(host, na_child_get_string(inst, "name")); - if (!(volume->volume_data.flags & VOLUME_INIT)) volume->volume_data.flags = volume_data->flags; - if (!(volume->volume_data.flags & VOLUME_DF)) continue; - size_free = na_child_get_uint64(inst, "size-available", 0); - size_used = na_child_get_uint64(inst, "size-used", 0); - snap_reserved = na_child_get_uint64(inst, "snapshot-blocks-reserved", 0) * 1024; - - vl.values_len = 1; - vl.values = values; - vl.time = time(0); - vl.interval = interval_g; - sstrncpy(vl.plugin, "netapp", sizeof(vl.plugin)); - sstrncpy(vl.host, host->name, sizeof(vl.host)); - sstrncpy(vl.plugin_instance, volume->name, sizeof(vl.plugin_instance)); - - values[0].gauge = size_used; - sstrncpy(vl.type_instance, "used", sizeof(vl.type_instance)); - DEBUG("%s/netapp-%s/df_complex-used: %llu", host->name, volume->name, size_used); - plugin_dispatch_values ("df_complex", &vl); - - values[0].gauge = size_free; - sstrncpy(vl.type_instance, "free", sizeof(vl.type_instance)); - DEBUG("%s/netapp-%s/df_complex-free: %llu", host->name, volume->name, size_free); - plugin_dispatch_values ("df_complex", &vl); - - if (snap_reserved) { - values[0].gauge = snap_reserved; - sstrncpy(vl.type_instance, "snap_reserved", sizeof(vl.type_instance)); - DEBUG("%s/netapp-%s/df_complex-snap_reserved: %llu", host->name, volume->name, snap_reserved); - plugin_dispatch_values ("df_complex", &vl); - } + uint64_t size_free = 0, size_used = 0, snap_reserved = 0; + + na_elem_t *sis; + const char *sis_state; + uint64_t sis_saved_reported; + uint64_t sis_saved; + + volume = get_volume(host, na_child_get_string(inst, "name"), + cfg_volume_data->flags, /* perf_flags = */ 0); + if (volume == NULL) + continue; + + if (!(volume->cfg_volume_usage.flags & VOLUME_DF)) + continue; + + /* 2^4 exa-bytes? This will take a while ;) */ + size_free = na_child_get_uint64(inst, "size-available", UINT64_MAX); + if (size_free != UINT64_MAX) + submit_double (host->name, volume->name, "df_complex", "used", + (double) size_used, /* time = */ 0); + + size_used = na_child_get_uint64(inst, "size-used", UINT64_MAX); + if (size_free != UINT64_MAX) + submit_double (host->name, volume->name, "df_complex", "free", + (double) size_free, /* time = */ 0); + + snap_reserved = na_child_get_uint64(inst, "snapshot-blocks-reserved", UINT64_MAX); + if (snap_reserved != UINT64_MAX) + /* 1 block == 1024 bytes as per API docs */ + submit_double (host->name, volume->name, "df_complex", "snap_reserved", + (double) (1024 * snap_reserved), /* time = */ 0); sis = na_elem_child(inst, "sis"); - if (sis && !strcmp(na_child_get_string(sis, "state"), "enabled")) { - uint64_t sis_saved_reported = na_child_get_uint64(sis, "size-saved", 0), sis_saved_percent = na_child_get_uint64(sis, "percentage-saved", 0); - /* size-saved is actually a 32 bit number, so ... time for some guesswork. */ - if (sis_saved_reported >> 32) { - /* In case they ever fix this bug. */ - sis_saved = sis_saved_reported; + if (sis == NULL) + continue; + + sis_state = na_child_get_string(sis, "state"); + if ((sis_state == NULL) + || (strcmp ("enabled", sis_state) != 0)) + continue; + + sis_saved_reported = na_child_get_uint64(sis, "size-saved", UINT64_MAX); + if (sis_saved_reported == UINT64_MAX) + continue; + + /* size-saved is actually a 32 bit number, so ... time for some guesswork. */ + if ((sis_saved_reported >> 32) != 0) { + /* In case they ever fix this bug. */ + sis_saved = sis_saved_reported; + } else { + uint64_t sis_saved_percent; + uint64_t sis_saved_guess; + uint64_t overflow_guess; + uint64_t guess1, guess2, guess3; + + sis_saved_percent = na_child_get_uint64(sis, "percentage-saved", UINT64_MAX); + if (sis_saved_percent > 100) + continue; + + /* The "size-saved" value is a 32bit unsigned integer. This is a bug and + * will hopefully be fixed in later versions. To work around the bug, try + * to figure out how often the 32bit integer wrapped around by using the + * "percentage-saved" value. Because the percentage is in the range + * [0-100], this should work as long as the saved space does not exceed + * 400 GBytes. */ + /* percentage-saved = size-saved / (size-saved + size-used) */ + if (sis_saved_percent < 100) + sis_saved_guess = size_used * sis_saved_percent / (100 - sis_saved_percent); + else + sis_saved_guess = size_used; + + overflow_guess = sis_saved_guess >> 32; + guess1 = overflow_guess ? ((overflow_guess - 1) << 32) + sis_saved_reported : sis_saved_reported; + guess2 = (overflow_guess << 32) + sis_saved_reported; + guess3 = ((overflow_guess + 1) << 32) + sis_saved_reported; + + if (sis_saved_guess < guess2) { + if ((sis_saved_guess - guess1) < (guess2 - sis_saved_guess)) + sis_saved = guess1; + else + sis_saved = guess2; } else { - uint64_t real_saved = sis_saved_percent * size_used / (100 - sis_saved_percent); - uint64_t overflow_guess = real_saved >> 32; - uint64_t guess1 = overflow_guess ? ((overflow_guess - 1) << 32) + sis_saved_reported : sis_saved_reported; - uint64_t guess2 = (overflow_guess << 32) + sis_saved_reported; - uint64_t guess3 = ((overflow_guess + 1) << 32) + sis_saved_reported; - - if (real_saved < guess2) { - if (real_saved - guess1 < guess2 - real_saved) sis_saved = guess1; - else sis_saved = guess2; - } else { - if (real_saved - guess2 < guess3 - real_saved) sis_saved = guess2; - else sis_saved = guess3; - } + if ((sis_saved_guess - guess2) < (guess3 - sis_saved_guess)) + sis_saved = guess2; + else + sis_saved = guess3; } - values[0].gauge = sis_saved; - sstrncpy(vl.type_instance, "sis_saved", sizeof(vl.type_instance)); - DEBUG("%s/netapp-%s/df_complex-sis_saved: %llu", host->name, volume->name, sis_saved); - plugin_dispatch_values ("df_complex", &vl); - } + } /* end of 32-bit workaround */ + + submit_double (host->name, volume->name, "df_complex", "sis_saved", + (double) sis_saved, /* time = */ 0); } -} +} /* }}} void collect_volume_data */ -static void collect_perf_volume_data(host_config_t *host, na_elem_t *out, void *data) { - perf_volume_data_t *perf = data; - const char *name; +/* Data corresponding to */ +static void query_volume_perf_data(host_config_t *host, na_elem_t *out, void *data) { /* {{{ */ + cfg_volume_perf_t *cfg_volume_perf = data; time_t timestamp; na_elem_t *counter, *inst; - volume_t *volume; - value_t values[2]; - value_list_t vl = VALUE_LIST_INIT; timestamp = (time_t) na_child_get_uint64(out, "timestamp", 0); + out = na_elem_child(out, "instances"); na_elem_iter_t inst_iter = na_child_iterator(out); for (inst = na_iterator_next(&inst_iter); inst; inst = na_iterator_next(&inst_iter)) { - uint64_t read_data = 0, write_data = 0, read_ops = 0, write_ops = 0, read_latency = 0, write_latency = 0; + data_volume_perf_t perf_data; + volume_t *volume; + + memset (&perf_data, 0, sizeof (perf_data)); + perf_data.timestamp = timestamp; + + volume = get_volume(host, na_child_get_string(inst, "name"), + /* data_flags = */ 0, cfg_volume_perf->flags); + if (volume == NULL) + continue; - volume = get_volume(host, na_child_get_string(inst, "name")); - if (!volume->perf_data.flags) { - volume->perf_data.flags = perf->flags; - volume->perf_data.last_read_latency = volume->perf_data.last_read_ops = 0; - volume->perf_data.last_write_latency = volume->perf_data.last_write_ops = 0; - } na_elem_iter_t count_iter = na_child_iterator(na_elem_child(inst, "counters")); for (counter = na_iterator_next(&count_iter); counter; counter = na_iterator_next(&count_iter)) { + const char *name; + uint64_t value; + name = na_child_get_string(counter, "name"); - if (!strcmp(name, "read_ops")) { - read_ops = na_child_get_uint64(counter, "value", 0); - } else if (!strcmp(name, "write_ops")) { - write_ops = na_child_get_uint64(counter, "value", 0); - } else if (!strcmp(name, "read_data")) { - read_data = na_child_get_uint64(counter, "value", 0); + if (name == NULL) + continue; + + value = na_child_get_uint64(counter, "value", UINT64_MAX); + if (value == UINT64_MAX) + continue; + + if (!strcmp(name, "read_data")) { + perf_data.read_bytes = value; + perf_data.flags |= HAVE_VOLUME_PERF_BYTES_READ; } else if (!strcmp(name, "write_data")) { - write_data = na_child_get_uint64(counter, "value", 0); + perf_data.write_bytes = value; + perf_data.flags |= HAVE_VOLUME_PERF_BYTES_WRITE; + } else if (!strcmp(name, "read_ops")) { + perf_data.read_ops = value; + perf_data.flags |= HAVE_VOLUME_PERF_OPS_READ; + } else if (!strcmp(name, "write_ops")) { + perf_data.write_ops = value; + perf_data.flags |= HAVE_VOLUME_PERF_OPS_WRITE; } else if (!strcmp(name, "read_latency")) { - read_latency = na_child_get_uint64(counter, "value", 0); + perf_data.read_latency = value; + perf_data.flags |= HAVE_VOLUME_PERF_LATENCY_READ; } else if (!strcmp(name, "write_latency")) { - write_latency = na_child_get_uint64(counter, "value", 0); - } - } - if (read_ops && write_ops) { - values[0].counter = read_ops; - values[1].counter = write_ops; - vl.values = values; - vl.values_len = 2; - vl.time = timestamp; - vl.interval = interval_g; - sstrncpy(vl.plugin, "netapp", sizeof(vl.plugin)); - sstrncpy(vl.host, host->name, sizeof(vl.host)); - sstrncpy(vl.plugin_instance, volume->name, sizeof(vl.plugin_instance)); - vl.type_instance[0] = 0; - if (volume->perf_data.flags & PERF_VOLUME_OPS) { - /* We might need the data even if it wasn't configured to calculate - the latency. Therefore we just skip the dispatch. */ - DEBUG("%s/netapp-%s/disk_ops: %llu %llu", host->name, instance, read_ops, write_ops); - plugin_dispatch_values("disk_ops", &vl); - } - if ((volume->perf_data.flags & PERF_VOLUME_LATENCY) && read_latency && write_latency) { - values[0].gauge = 0; - if (read_ops - volume->perf_data.last_read_ops) values[0].gauge = (read_latency - volume->perf_data.last_read_latency) * (timestamp - volume->perf_data.last_timestamp) / (read_ops - volume->perf_data.last_read_ops); - values[1].gauge = 0; - if (write_ops - volume->perf_data.last_write_ops) values[1].gauge = (write_latency - volume->perf_data.last_write_latency) * (timestamp - volume->perf_data.last_timestamp) / (write_ops - volume->perf_data.last_write_ops); - vl.values = values; - vl.values_len = 2; - vl.time = timestamp; - vl.interval = interval_g; - sstrncpy(vl.plugin, "netapp", sizeof(vl.plugin)); - sstrncpy(vl.host, host->name, sizeof(vl.host)); - sstrncpy(vl.plugin_instance, volume->name, sizeof(vl.plugin_instance)); - vl.type_instance[0] = 0; - if (volume->perf_data.last_read_ops && volume->perf_data.last_write_ops) { - DEBUG("%s/netapp-%s/disk_latency: lrlc: %llu ro: %llu lro: %llu rl: %llu lrl: %llu %llu %llu", host->name, instance, volume->perf_data.last_read_latency_counter, read_ops, volume->perf_data.last_read_ops, read_latency, volume->perf_data.last_read_latency, values[0].counter, values[1].counter); - plugin_dispatch_values("disk_latency", &vl); - } - volume->perf_data.last_timestamp = timestamp; - volume->perf_data.last_read_latency = read_latency; - volume->perf_data.last_read_ops = read_ops; - volume->perf_data.last_write_latency = write_latency; - volume->perf_data.last_write_ops = write_ops; + perf_data.write_latency = value; + perf_data.flags |= HAVE_VOLUME_PERF_LATENCY_WRITE; } } - if ((volume->perf_data.flags & PERF_VOLUME_IO) && read_data && write_data) { - values[0].counter = read_data; - values[1].counter = write_data; - vl.values = values; - vl.values_len = 2; - vl.time = timestamp; - vl.interval = interval_g; - sstrncpy(vl.plugin, "netapp", sizeof(vl.plugin)); - sstrncpy(vl.host, host->name, sizeof(vl.host)); - sstrncpy(vl.plugin_instance, volume->name, sizeof(vl.plugin_instance)); - vl.type_instance[0] = 0; - DEBUG("%s/netapp-%s/disk_octets: %llu %llu", host->name, instance, read_data, write_data); - plugin_dispatch_values ("disk_octets", &vl); - } - } -} -static void collect_perf_system_data(host_config_t *host, na_elem_t *out, void *data) { - uint64_t disk_read = 0, disk_written = 0, net_recv = 0, net_sent = 0, cpu_busy = 0, cpu_total = 0; - perf_system_data_t *perf = data; - const char *instance, *name; + submit_volume_perf_data (host, volume, &perf_data); + } /* for (volume) */ +} /* }}} void query_volume_perf_data */ + +/* Data corresponding to */ +static void collect_perf_system_data(host_config_t *host, na_elem_t *out, void *data) { /* {{{ */ + counter_t disk_read = 0, disk_written = 0; + counter_t net_recv = 0, net_sent = 0; + counter_t cpu_busy = 0, cpu_total = 0; + unsigned int counter_flags = 0; + + cfg_system_t *cfg_system = data; + const char *instance; time_t timestamp; na_elem_t *counter; - value_t values[2]; - value_list_t vl = VALUE_LIST_INIT; timestamp = (time_t) na_child_get_uint64(out, "timestamp", 0); out = na_elem_child(na_elem_child(out, "instances"), "instance-data"); @@ -544,489 +911,410 @@ static void collect_perf_system_data(host_config_t *host, na_elem_t *out, void * na_elem_iter_t iter = na_child_iterator(na_elem_child(out, "counters")); for (counter = na_iterator_next(&iter); counter; counter = na_iterator_next(&iter)) { + const char *name; + uint64_t value; + name = na_child_get_string(counter, "name"); + if (name == NULL) + continue; + + value = na_child_get_uint64(counter, "value", UINT64_MAX); + if (value == UINT64_MAX) + continue; + if (!strcmp(name, "disk_data_read")) { - disk_read = na_child_get_uint64(counter, "value", 0) * 1024; + disk_read = (counter_t) (value * 1024); + counter_flags |= 0x01; } else if (!strcmp(name, "disk_data_written")) { - disk_written = na_child_get_uint64(counter, "value", 0) * 1024; + disk_written = (counter_t) (value * 1024); + counter_flags |= 0x02; } else if (!strcmp(name, "net_data_recv")) { - net_recv = na_child_get_uint64(counter, "value", 0) * 1024; + net_recv = (counter_t) (value * 1024); + counter_flags |= 0x04; } else if (!strcmp(name, "net_data_sent")) { - net_sent = na_child_get_uint64(counter, "value", 0) * 1024; + net_sent = (counter_t) (value * 1024); + counter_flags |= 0x08; } else if (!strcmp(name, "cpu_busy")) { - cpu_busy = na_child_get_uint64(counter, "value", 0); + cpu_busy = (counter_t) value; + counter_flags |= 0x10; } else if (!strcmp(name, "cpu_elapsed_time")) { - cpu_total = na_child_get_uint64(counter, "value", 0); - } else if ((perf->flags & PERF_SYSTEM_OPS) && strlen(name) > 4 && !strcmp(name + strlen(name) - 4, "_ops")) { - values[0].counter = na_child_get_uint64(counter, "value", 0); - if (!values[0].counter) continue; - vl.values = values; - vl.values_len = 1; - vl.time = timestamp; - vl.interval = interval_g; - sstrncpy(vl.plugin, "netapp", sizeof(vl.plugin)); - sstrncpy(vl.host, host->name, sizeof(vl.host)); - sstrncpy(vl.plugin_instance, instance, sizeof(vl.plugin_instance)); - sstrncpy(vl.type_instance, name, sizeof(vl.plugin_instance)); - DEBUG("%s/netapp-%s/disk_ops_complex-%s: %llu", host->name, instance, name, values[0].counter); - plugin_dispatch_values ("disk_ops_complex", &vl); + cpu_total = (counter_t) value; + counter_flags |= 0x20; + } else if ((cfg_system->flags & CFG_SYSTEM_OPS) + && (strlen(name) > 4) + && (!strcmp(name + strlen(name) - 4, "_ops"))) { + submit_counter (host->name, instance, "disk_ops_complex", name, + (counter_t) value, timestamp); } - } - if ((perf->flags & PERF_SYSTEM_DISK) && disk_read && disk_written) { - values[0].counter = disk_read; - values[1].counter = disk_written; - vl.values = values; - vl.values_len = 2; - vl.time = timestamp; - vl.interval = interval_g; - sstrncpy(vl.plugin, "netapp", sizeof(vl.plugin)); - sstrncpy(vl.host, host->name, sizeof(vl.host)); - sstrncpy(vl.plugin_instance, instance, sizeof(vl.plugin_instance)); - vl.type_instance[0] = 0; - DEBUG("%s/netapp-%s/disk_octets: %llu %llu", host->name, instance, disk_read, disk_written); - plugin_dispatch_values ("disk_octets", &vl); - } - if ((perf->flags & PERF_SYSTEM_NET) && net_recv && net_sent) { - values[0].counter = net_recv; - values[1].counter = net_sent; - vl.values = values; - vl.values_len = 2; - vl.time = timestamp; - vl.interval = interval_g; - sstrncpy(vl.plugin, "netapp", sizeof(vl.plugin)); - sstrncpy(vl.host, host->name, sizeof(vl.host)); - sstrncpy(vl.plugin_instance, instance, sizeof(vl.plugin_instance)); - vl.type_instance[0] = 0; - DEBUG("%s/netapp-%s/if_octects: %llu %llu", host->name, instance, net_recv, net_sent); - plugin_dispatch_values ("if_octets", &vl); - } - if ((perf->flags & PERF_SYSTEM_CPU) && cpu_busy && cpu_total) { -// values[0].gauge = (double) (cpu_busy - perf->last_cpu_busy) / (cpu_total - perf->last_cpu_total) * 100; - values[0].counter = cpu_busy / 10000; - vl.values = values; - vl.values_len = 1; - vl.time = timestamp; - vl.interval = interval_g; - sstrncpy(vl.plugin, "netapp", sizeof(vl.plugin)); - sstrncpy(vl.host, host->name, sizeof(vl.host)); - sstrncpy(vl.plugin_instance, instance, sizeof(vl.plugin_instance)); - sstrncpy(vl.type_instance, "system", sizeof(vl.plugin_instance)); -// if (perf->last_cpu_busy && perf->last_cpu_total) printf("CPU: busy: %lf - idle: %lf\n", values[0].gauge, 100.0 - values[0].gauge); -// if (perf->last_cpu_busy && perf->last_cpu_total) plugin_dispatch_values ("cpu", &vl); - DEBUG("%s/netapp-%s/cpu: busy: %llu - idle: %llu", host->name, instance, cpu_busy / 10000, cpu_total / 10000); - plugin_dispatch_values ("cpu", &vl); - -// values[0].gauge = 100.0 - (double) (cpu_busy - perf->last_cpu_busy) / (cpu_total - perf->last_cpu_total) * 100; - values[0].counter = (cpu_total - cpu_busy) / 10000; - vl.values = values; - vl.values_len = 1; - vl.time = timestamp; - vl.interval = interval_g; - sstrncpy(vl.plugin, "netapp", sizeof(vl.plugin)); - sstrncpy(vl.host, host->name, sizeof(vl.host)); - sstrncpy(vl.plugin_instance, instance, sizeof(vl.plugin_instance)); - sstrncpy(vl.type_instance, "idle", sizeof(vl.plugin_instance)); -// if (perf->last_cpu_busy && perf->last_cpu_total) plugin_dispatch_values ("cpu", &vl); - plugin_dispatch_values ("cpu", &vl); - - perf->last_cpu_busy = cpu_busy; - perf->last_cpu_total = cpu_total; - } -} + } /* for (counter) */ -int config_init() { - char err[256]; - na_elem_t *e; - host_config_t *host; - service_config_t *service; - - if (!host_config) { - WARNING("netapp plugin: Plugin loaded but no hosts defined."); - return 1; + if ((cfg_system->flags & CFG_SYSTEM_DISK) + && ((counter_flags & 0x03) == 0x03)) + submit_two_counters (host->name, instance, "disk_octets", NULL, + disk_read, disk_written, timestamp); + + if ((cfg_system->flags & CFG_SYSTEM_NET) + && ((counter_flags & 0x0c) == 0x0c)) + submit_two_counters (host->name, instance, "if_octets", NULL, + net_recv, net_sent, timestamp); + + if ((cfg_system->flags & CFG_SYSTEM_CPU) + && ((counter_flags & 0x30) == 0x30)) { + submit_counter (host->name, instance, "cpu", "system", + cpu_busy, timestamp); + submit_counter (host->name, instance, "cpu", "idle", + cpu_total - cpu_busy, timestamp); } +} /* }}} void collect_perf_system_data */ - if (!na_startup(err, sizeof(err))) { - ERROR("netapp plugin: Error initializing netapp API: %s", err); - return 1; +/* + * Configuration handling + */ +/* Sets a given flag if the boolean argument is true and unsets the flag if it + * is false. On error, the flag-field is not changed. */ +static int cna_config_bool_to_flag (const oconfig_item_t *ci, /* {{{ */ + uint32_t *flags, uint32_t flag) +{ + if ((ci == NULL) || (flags == NULL)) + return (EINVAL); + + if ((ci->values_num != 1) || (ci->values[0].type != OCONFIG_TYPE_BOOLEAN)) + { + WARNING ("netapp plugin: The %s option needs exactly one boolean argument.", + ci->key); + return (-1); } - for (host = host_config; host; host = host->next) { - host->srv = na_server_open(host->host, 1, 1); - na_server_set_transport_type(host->srv, host->protocol, 0); - na_server_set_port(host->srv, host->port); - na_server_style(host->srv, NA_STYLE_LOGIN_PASSWORD); - na_server_adminuser(host->srv, host->username, host->password); - na_server_set_timeout(host->srv, 5); - for (service = host->services; service; service = service->next) { - service->interval = host->interval * service->multiplier; - if (service->handler == collect_perf_system_data) { - service->query = na_elem_new("perf-object-get-instances"); - na_child_add_string(service->query, "objectname", "system"); - } else if (service->handler == collect_perf_volume_data) { - service->query = na_elem_new("perf-object-get-instances"); - na_child_add_string(service->query, "objectname", "volume"); -/* e = na_elem_new("instances"); - na_child_add_string(e, "foo", "system"); - na_child_add(root, e);*/ - e = na_elem_new("counters"); - na_child_add_string(e, "foo", "read_ops"); - na_child_add_string(e, "foo", "write_ops"); - na_child_add_string(e, "foo", "read_data"); - na_child_add_string(e, "foo", "write_data"); - na_child_add_string(e, "foo", "read_latency"); - na_child_add_string(e, "foo", "write_latency"); - na_child_add(service->query, e); - } else if (service->handler == collect_perf_wafl_data) { - service->query = na_elem_new("perf-object-get-instances"); - na_child_add_string(service->query, "objectname", "wafl"); -/* e = na_elem_new("instances"); - na_child_add_string(e, "foo", "system"); - na_child_add(root, e);*/ - e = na_elem_new("counters"); - na_child_add_string(e, "foo", "name_cache_hit"); - na_child_add_string(e, "foo", "name_cache_miss"); - na_child_add_string(e, "foo", "find_dir_hit"); - na_child_add_string(e, "foo", "find_dir_miss"); - na_child_add_string(e, "foo", "buf_hash_hit"); - na_child_add_string(e, "foo", "buf_hash_miss"); - na_child_add_string(e, "foo", "inode_cache_hit"); - na_child_add_string(e, "foo", "inode_cache_miss"); -// na_child_add_string(e, "foo", "inode_eject_time"); -// na_child_add_string(e, "foo", "buf_eject_time"); - na_child_add(service->query, e); - } else if (service->handler == collect_perf_disk_data) { - service->query = na_elem_new("perf-object-get-instances"); - na_child_add_string(service->query, "objectname", "disk"); - e = na_elem_new("counters"); - na_child_add_string(e, "foo", "disk_busy"); - na_child_add_string(e, "foo", "base_for_disk_busy"); - na_child_add(service->query, e); - } else if (service->handler == collect_volume_data) { - service->query = na_elem_new("volume-list-info"); -// na_child_add_string(service->query, "objectname", "volume"); -// } else if (service->handler == collect_snapshot_data) { -// service->query = na_elem_new("snapshot-list-info"); - } - } - } - return 0; -} + if (ci->values[0].value.boolean) + *flags |= flag; + else + *flags &= ~flag; -static void set_global_perf_vol_flag(const host_config_t *host, uint32_t flag, int value) { - volume_t *v; - - for (v = host->volumes; v; v = v->next) { - v->perf_data.flags &= ~flag; - if (value) v->perf_data.flags |= flag; + return (0); +} /* }}} int cna_config_bool_to_flag */ + +/* Handling of the "Multiplier" option which is allowed in every block. */ +static int cna_config_get_multiplier (const oconfig_item_t *ci, /* {{{ */ + cfg_service_t *service) +{ + int tmp; + + if ((ci == NULL) || (service == NULL)) + return (EINVAL); + + if ((ci->values_num != 1) || (ci->values[0].type != OCONFIG_TYPE_NUMBER)) + { + WARNING ("netapp plugin: The `Multiplier' option needs exactly one numeric argument."); + return (-1); } -} -static void set_global_vol_flag(const host_config_t *host, uint32_t flag, int value) { - volume_t *v; - - for (v = host->volumes; v; v = v->next) { - v->volume_data.flags &= ~flag; - if (value) v->volume_data.flags |= flag; + tmp = (int) (ci->values[0].value.number + .5); + if (tmp < 1) + { + WARNING ("netapp plugin: The `Multiplier' option needs a positive integer argument."); + return (-1); } -} -static void process_perf_volume_flag(host_config_t *host, perf_volume_data_t *perf_volume, const oconfig_item_t *item, uint32_t flag) { - int n; + service->multiplier = tmp; + service->skip_countdown = tmp; + + return (0); +} /* }}} int cna_config_get_multiplier */ + +/* Handling of the "GetIO", "GetOps" and "GetLatency" options within a + * block. */ +static void cna_config_volume_performance_option (host_config_t *host, /* {{{ */ + cfg_volume_perf_t *perf_volume, const oconfig_item_t *item, + uint32_t flag) +{ + int i; - for (n = 0; n < item->values_num; ++n) { - int minus = 0; - const char *name = item->values[n].value.string; + for (i = 0; i < item->values_num; ++i) { + const char *name; volume_t *v; - if (item->values[n].type != OCONFIG_TYPE_STRING) { - WARNING("netapp plugin: Ignoring non-string argument in \"GetVolPerfData\" block for host %s", host->name); + _Bool set = true; + + if (item->values[i].type != OCONFIG_TYPE_STRING) { + WARNING("netapp plugin: Ignoring non-string argument in " + "\"GetVolumePerfData\" block for host %s", host->name); continue; } + + name = item->values[i].value.string; if (name[0] == '+') { + set = true; ++name; } else if (name[0] == '-') { - minus = 1; + set = false; ++name; } + if (!name[0]) { - perf_volume->flags &= ~flag; - if (!minus) perf_volume->flags |= flag; - set_global_perf_vol_flag(host, flag, !minus); - continue; - } - v = get_volume(host, name); - if (!v->perf_data.flags) { - v->perf_data.flags = perf_volume->flags; - v->perf_data.last_read_latency = v->perf_data.last_read_ops = 0; - v->perf_data.last_write_latency = v->perf_data.last_write_ops = 0; - } - v->perf_data.flags &= ~flag; - if (!minus) v->perf_data.flags |= flag; - } -} + if (set) + perf_volume->flags |= flag; + else /* if (!set) */ + perf_volume->flags &= ~flag; -static void process_volume_flag(host_config_t *host, volume_data_t *volume_data, const oconfig_item_t *item, uint32_t flag) { - int n; - - for (n = 0; n < item->values_num; ++n) { - int minus = 0; - const char *name = item->values[n].value.string; - volume_t *v; - if (item->values[n].type != OCONFIG_TYPE_STRING) { - WARNING("netapp plugin: Ignoring non-string argument in \"GetVolData\" block for host %s", host->name); + set_global_perf_vol_flag(host, flag, set); continue; } - if (name[0] == '+') { - ++name; - } else if (name[0] == '-') { - minus = 1; - ++name; - } - if (!name[0]) { - volume_data->flags &= ~flag; - if (!minus) volume_data->flags |= flag; - set_global_vol_flag(host, flag, !minus); + + v = get_volume (host, name, /* data_flags = */ 0, perf_volume->flags); + if (v == NULL) continue; - } - v = get_volume(host, name); - if (!v->volume_data.flags) v->volume_data.flags = volume_data->flags; - v->volume_data.flags &= ~flag; - if (!minus) v->volume_data.flags |= flag; - } -} -static void build_perf_vol_config(host_config_t *host, const oconfig_item_t *ci) { + if (set) + v->perf_data.flags |= flag; + else /* if (!set) */ + v->perf_data.flags &= ~flag; + } /* for (i = 0 .. item->values_num) */ +} /* }}} void cna_config_volume_performance_option */ + +/* Corresponds to a block */ +static void cna_config_volume_performance(host_config_t *host, const oconfig_item_t *ci) { /* {{{ */ int i, had_io = 0, had_ops = 0, had_latency = 0; - service_config_t *service; - perf_volume_data_t *perf_volume; + cfg_service_t *service; + cfg_volume_perf_t *perf_volume; service = malloc(sizeof(*service)); service->query = 0; - service->handler = collect_perf_volume_data; + service->handler = query_volume_perf_data; perf_volume = service->data = malloc(sizeof(*perf_volume)); - perf_volume->flags = PERF_VOLUME_INIT; + perf_volume->flags = CFG_VOLUME_PERF_INIT; service->next = host->services; host->services = service; for (i = 0; i < ci->children_num; ++i) { oconfig_item_t *item = ci->children + i; -// if (!item || !item->key || !*item->key) continue; + /* if (!item || !item->key || !*item->key) continue; */ if (!strcasecmp(item->key, "Multiplier")) { - if (item->values_num != 1 || item->values[0].type != OCONFIG_TYPE_NUMBER || item->values[0].value.number != (int) item->values[0].value.number || item->values[0].value.number < 1) { - WARNING("netapp plugin: \"Multiplier\" of host %s service GetVolPerfData needs exactly one positive integer argument.", host->name); - continue; - } - service->skip_countdown = service->multiplier = item->values[0].value.number; + cna_config_get_multiplier (item, service); } else if (!strcasecmp(item->key, "GetIO")) { had_io = 1; - process_perf_volume_flag(host, perf_volume, item, PERF_VOLUME_IO); + cna_config_volume_performance_option(host, perf_volume, item, CFG_VOLUME_PERF_IO); } else if (!strcasecmp(item->key, "GetOps")) { had_ops = 1; - process_perf_volume_flag(host, perf_volume, item, PERF_VOLUME_OPS); + cna_config_volume_performance_option(host, perf_volume, item, CFG_VOLUME_PERF_OPS); } else if (!strcasecmp(item->key, "GetLatency")) { had_latency = 1; - process_perf_volume_flag(host, perf_volume, item, PERF_VOLUME_LATENCY); + cna_config_volume_performance_option(host, perf_volume, item, CFG_VOLUME_PERF_LATENCY); } } if (!had_io) { - perf_volume->flags |= PERF_VOLUME_IO; - set_global_perf_vol_flag(host, PERF_VOLUME_IO, 1); + perf_volume->flags |= CFG_VOLUME_PERF_IO; + set_global_perf_vol_flag(host, CFG_VOLUME_PERF_IO, /* set = */ true); } if (!had_ops) { - perf_volume->flags |= PERF_VOLUME_OPS; - set_global_perf_vol_flag(host, PERF_VOLUME_OPS, 1); + perf_volume->flags |= CFG_VOLUME_PERF_OPS; + set_global_perf_vol_flag(host, CFG_VOLUME_PERF_OPS, /* set = */ true); } if (!had_latency) { - perf_volume->flags |= PERF_VOLUME_LATENCY; - set_global_perf_vol_flag(host, PERF_VOLUME_LATENCY, 1); + perf_volume->flags |= CFG_VOLUME_PERF_LATENCY; + set_global_perf_vol_flag(host, CFG_VOLUME_PERF_LATENCY, /* set = */ true); } -} +} /* }}} void cna_config_volume_performance */ -static void build_volume_config(host_config_t *host, oconfig_item_t *ci) { +/* Handling of the "GetDiskUtil" option within a block. */ +static void cna_config_volume_usage_option (host_config_t *host, /* {{{ */ + cfg_volume_usage_t *cfg_volume_data, const oconfig_item_t *item, uint32_t flag) +{ + int i; + + for (i = 0; i < item->values_num; ++i) { + const char *name; + volume_t *v; + _Bool set = true; + + if (item->values[i].type != OCONFIG_TYPE_STRING) { + WARNING("netapp plugin: Ignoring non-string argument in \"GetVolData\"" + "block for host %s", host->name); + continue; + } + + name = item->values[i].value.string; + if (name[0] == '+') { + set = true; + ++name; + } else if (name[0] == '-') { + set = false; + ++name; + } + + if (!name[0]) { + if (set) + cfg_volume_data->flags |= flag; + else /* if (!set) */ + cfg_volume_data->flags &= ~flag; + + set_global_vol_flag(host, flag, set); + continue; + } + + v = get_volume(host, name, cfg_volume_data->flags, /* perf_flags = */ 0); + if (v == NULL) + continue; + + if (!v->cfg_volume_usage.flags) + v->cfg_volume_usage.flags = cfg_volume_data->flags; + + if (set) + v->cfg_volume_usage.flags |= flag; + else /* if (!set) */ + v->cfg_volume_usage.flags &= ~flag; + } +} /* }}} void cna_config_volume_usage_option */ + +/* Corresponds to a block */ +static void cna_config_volume_usage(host_config_t *host, oconfig_item_t *ci) { /* {{{ */ int i, had_df = 0; - service_config_t *service; - volume_data_t *volume_data; + cfg_service_t *service; + cfg_volume_usage_t *cfg_volume_data; service = malloc(sizeof(*service)); service->query = 0; service->handler = collect_volume_data; - volume_data = service->data = malloc(sizeof(*volume_data)); - volume_data->flags = VOLUME_INIT; + cfg_volume_data = service->data = malloc(sizeof(*cfg_volume_data)); + cfg_volume_data->flags = VOLUME_INIT; service->next = host->services; host->services = service; for (i = 0; i < ci->children_num; ++i) { oconfig_item_t *item = ci->children + i; -// if (!item || !item->key || !*item->key) continue; + /* if (!item || !item->key || !*item->key) continue; */ if (!strcasecmp(item->key, "Multiplier")) { - if (item->values_num != 1 || item->values[0].type != OCONFIG_TYPE_NUMBER || item->values[0].value.number != (int) item->values[0].value.number || item->values[0].value.number < 1) { - WARNING("netapp plugin: \"Multiplier\" of host %s service GetVolPerfData needs exactly one positive integer argument.", host->name); - continue; - } - service->skip_countdown = service->multiplier = item->values[0].value.number; + cna_config_get_multiplier (item, service); } else if (!strcasecmp(item->key, "GetDiskUtil")) { had_df = 1; - process_volume_flag(host, volume_data, item, VOLUME_DF); + cna_config_volume_usage_option(host, cfg_volume_data, item, VOLUME_DF); } } if (!had_df) { - volume_data->flags |= VOLUME_DF; - set_global_vol_flag(host, VOLUME_DF, 1); + cfg_volume_data->flags |= VOLUME_DF; + set_global_vol_flag(host, VOLUME_DF, /* set = */ true); } -/* service = malloc(sizeof(*service)); - service->query = 0; - service->handler = collect_snapshot_data; - service->data = volume_data; - service->next = temp->services; - temp->services = service;*/ -} +} /* }}} void cna_config_volume_usage */ -static void build_perf_disk_config(host_config_t *temp, oconfig_item_t *ci) { +/* Corresponds to a block */ +static void cna_config_disk(host_config_t *temp, oconfig_item_t *ci) { /* {{{ */ int i; - service_config_t *service; - perf_disk_data_t *perf_disk; + cfg_service_t *service; + cfg_disk_t *cfg_disk; service = malloc(sizeof(*service)); service->query = 0; - service->handler = collect_perf_disk_data; - perf_disk = service->data = malloc(sizeof(*perf_disk)); - perf_disk->flags = PERF_DISK_ALL; + service->handler = query_submit_disk_data; + cfg_disk = service->data = malloc(sizeof(*cfg_disk)); + cfg_disk->flags = CFG_DISK_ALL; service->next = temp->services; temp->services = service; for (i = 0; i < ci->children_num; ++i) { oconfig_item_t *item = ci->children + i; -// if (!item || !item->key || !*item->key) continue; + /* if (!item || !item->key || !*item->key) continue; */ if (!strcasecmp(item->key, "Multiplier")) { - if (item->values_num != 1 || item->values[0].type != OCONFIG_TYPE_NUMBER || item->values[0].value.number != (int) item->values[0].value.number || item->values[0].value.number < 1) { - WARNING("netapp plugin: \"Multiplier\" of host %s service GetWaflPerfData needs exactly one positive integer argument.", ci->values[0].value.string); - continue; - } - service->skip_countdown = service->multiplier = item->values[0].value.number; + cna_config_get_multiplier (item, service); } else if (!strcasecmp(item->key, "GetBusy")) { - if (item->values_num != 1 || item->values[0].type != OCONFIG_TYPE_BOOLEAN) { - WARNING("netapp plugin: \"GetBusy\" of host %s service GetDiskPerfData needs exactly one bool argument.", ci->values[0].value.string); - continue; - } - perf_disk->flags = (perf_disk->flags & ~PERF_SYSTEM_CPU) | (item->values[0].value.boolean ? PERF_DISK_BUSIEST : 0); + cna_config_bool_to_flag (item, &cfg_disk->flags, CFG_SYSTEM_CPU); } } -} +} /* }}} void cna_config_disk */ -static void build_perf_wafl_config(host_config_t *temp, oconfig_item_t *ci) { +/* Corresponds to a block */ +static void cna_config_wafl(host_config_t *host, oconfig_item_t *ci) { /* {{{ */ int i; - service_config_t *service; - perf_wafl_data_t *perf_wafl; + cfg_service_t *service; + data_wafl_t *perf_wafl; service = malloc(sizeof(*service)); + if (service == NULL) + return; + memset (service, 0, sizeof (*service)); + service->query = 0; - service->handler = collect_perf_wafl_data; + service->handler = query_wafl_data; perf_wafl = service->data = malloc(sizeof(*perf_wafl)); - perf_wafl->flags = PERF_WAFL_ALL; - perf_wafl->last_name_cache_hit = 0; - perf_wafl->last_name_cache_miss = 0; - perf_wafl->last_find_dir_hit = 0; - perf_wafl->last_find_dir_miss = 0; - perf_wafl->last_buf_hash_hit = 0; - perf_wafl->last_buf_hash_miss = 0; - perf_wafl->last_inode_cache_hit = 0; - perf_wafl->last_inode_cache_miss = 0; - service->next = temp->services; - temp->services = service; + perf_wafl->flags = CFG_WAFL_ALL; + for (i = 0; i < ci->children_num; ++i) { oconfig_item_t *item = ci->children + i; -// if (!item || !item->key || !*item->key) continue; if (!strcasecmp(item->key, "Multiplier")) { - if (item->values_num != 1 || item->values[0].type != OCONFIG_TYPE_NUMBER || item->values[0].value.number != (int) item->values[0].value.number || item->values[0].value.number < 1) { - WARNING("netapp plugin: \"Multiplier\" of host %s service GetWaflPerfData needs exactly one positive integer argument.", ci->values[0].value.string); - continue; - } - service->skip_countdown = service->multiplier = item->values[0].value.number; + cna_config_get_multiplier (item, service); } else if (!strcasecmp(item->key, "GetNameCache")) { - if (item->values_num != 1 || item->values[0].type != OCONFIG_TYPE_BOOLEAN) { - WARNING("netapp plugin: \"GetNameCache\" of host %s service GetWaflPerfData needs exactly one bool argument.", ci->values[0].value.string); - continue; - } - perf_wafl->flags = (perf_wafl->flags & ~PERF_WAFL_NAME_CACHE) | (item->values[0].value.boolean ? PERF_WAFL_NAME_CACHE : 0); + cna_config_bool_to_flag (item, &perf_wafl->flags, CFG_WAFL_NAME_CACHE); } else if (!strcasecmp(item->key, "GetDirCache")) { - if (item->values_num != 1 || item->values[0].type != OCONFIG_TYPE_BOOLEAN) { - WARNING("netapp plugin: \"GetDirChache\" of host %s service GetWaflPerfData needs exactly one bool argument.", ci->values[0].value.string); - continue; - } - perf_wafl->flags = (perf_wafl->flags & ~PERF_WAFL_DIR_CACHE) | (item->values[0].value.boolean ? PERF_WAFL_DIR_CACHE : 0); + cna_config_bool_to_flag (item, &perf_wafl->flags, CFG_WAFL_DIR_CACHE); } else if (!strcasecmp(item->key, "GetBufCache")) { - if (item->values_num != 1 || item->values[0].type != OCONFIG_TYPE_BOOLEAN) { - WARNING("netapp plugin: \"GetBufCache\" of host %s service GetWaflPerfData needs exactly one bool argument.", ci->values[0].value.string); - continue; - } - perf_wafl->flags = (perf_wafl->flags & ~PERF_WAFL_BUF_CACHE) | (item->values[0].value.boolean ? PERF_WAFL_BUF_CACHE : 0); + cna_config_bool_to_flag (item, &perf_wafl->flags, CFG_WAFL_BUF_CACHE); } else if (!strcasecmp(item->key, "GetInodeCache")) { - if (item->values_num != 1 || item->values[0].type != OCONFIG_TYPE_BOOLEAN) { - WARNING("netapp plugin: \"GetInodeCache\" of host %s service GetWaflPerfData needs exactly one bool argument.", ci->values[0].value.string); - continue; - } - perf_wafl->flags = (perf_wafl->flags & ~PERF_WAFL_INODE_CACHE) | (item->values[0].value.boolean ? PERF_WAFL_INODE_CACHE : 0); + cna_config_bool_to_flag (item, &perf_wafl->flags, CFG_WAFL_INODE_CACHE); + } else { + WARNING ("netapp plugin: The %s config option is not allowed within " + "`GetWaflPerfData' blocks.", item->key); } } -} -static void build_perf_sys_config(host_config_t *temp, oconfig_item_t *ci, const service_config_t *default_service) { + service->next = host->services; + host->services = service; +} /* }}} void cna_config_wafl */ + +/* Corresponds to a block */ +static int cna_config_system (host_config_t *host, /* {{{ */ + oconfig_item_t *ci, const cfg_service_t *default_service) +{ int i; - service_config_t *service; - perf_system_data_t *perf_system; + cfg_service_t *service; + cfg_system_t *cfg_system; service = malloc(sizeof(*service)); + if (service == NULL) + return (-1); + memset (service, 0, sizeof (*service)); *service = *default_service; service->handler = collect_perf_system_data; - perf_system = service->data = malloc(sizeof(*perf_system)); - perf_system->flags = PERF_SYSTEM_ALL; - perf_system->last_cpu_busy = 0; - perf_system->last_cpu_total = 0; - service->next = temp->services; - temp->services = service; + + cfg_system = malloc(sizeof(*cfg_system)); + if (cfg_system == NULL) { + sfree (service); + return (-1); + } + memset (cfg_system, 0, sizeof (*cfg_system)); + cfg_system->flags = CFG_SYSTEM_ALL; + service->data = cfg_system; + for (i = 0; i < ci->children_num; ++i) { oconfig_item_t *item = ci->children + i; -// if (!item || !item->key || !*item->key) continue; if (!strcasecmp(item->key, "Multiplier")) { - if (item->values_num != 1 || item->values[0].type != OCONFIG_TYPE_NUMBER || item->values[0].value.number != (int) item->values[0].value.number || item->values[0].value.number < 1) { - WARNING("netapp plugin: \"Multiplier\" of host %s service GetSystemPerfData needs exactly one positive integer argument.", ci->values[0].value.string); - continue; - } - service->skip_countdown = service->multiplier = item->values[0].value.number; + cna_config_get_multiplier (item, service); } else if (!strcasecmp(item->key, "GetCPULoad")) { - if (item->values_num != 1 || item->values[0].type != OCONFIG_TYPE_BOOLEAN) { - WARNING("netapp plugin: \"GetCPULoad\" of host %s service GetSystemPerfData needs exactly one bool argument.", ci->values[0].value.string); - continue; - } - perf_system->flags = (perf_system->flags & ~PERF_SYSTEM_CPU) | (item->values[0].value.boolean ? PERF_SYSTEM_CPU : 0); + cna_config_bool_to_flag (item, &cfg_system->flags, CFG_SYSTEM_CPU); } else if (!strcasecmp(item->key, "GetInterfaces")) { - if (item->values_num != 1 || item->values[0].type != OCONFIG_TYPE_BOOLEAN) { - WARNING("netapp plugin: \"GetInterfaces\" of host %s service GetSystemPerfData needs exactly one bool argument.", ci->values[0].value.string); - continue; - } - perf_system->flags = (perf_system->flags & ~PERF_SYSTEM_NET) | (item->values[0].value.boolean ? PERF_SYSTEM_NET : 0); + cna_config_bool_to_flag (item, &cfg_system->flags, CFG_SYSTEM_NET); } else if (!strcasecmp(item->key, "GetDiskOps")) { - if (item->values_num != 1 || item->values[0].type != OCONFIG_TYPE_BOOLEAN) { - WARNING("netapp plugin: \"GetDiskOps\" of host %s service GetSystemPerfData needs exactly one bool argument.", ci->values[0].value.string); - continue; - } - perf_system->flags = (perf_system->flags & ~PERF_SYSTEM_OPS) | (item->values[0].value.boolean ? PERF_SYSTEM_OPS : 0); + cna_config_bool_to_flag (item, &cfg_system->flags, CFG_SYSTEM_OPS); } else if (!strcasecmp(item->key, "GetDiskIO")) { - if (item->values_num != 1 || item->values[0].type != OCONFIG_TYPE_BOOLEAN) { - WARNING("netapp plugin: \"GetDiskIO\" of host %s service GetSystemPerfData needs exactly one bool argument.", ci->values[0].value.string); - continue; - } - perf_system->flags = (perf_system->flags & ~PERF_SYSTEM_DISK) | (item->values[0].value.boolean ? PERF_SYSTEM_DISK : 0); + cna_config_bool_to_flag (item, &cfg_system->flags, CFG_SYSTEM_DISK); + } else { + WARNING ("netapp plugin: The %s config option is not allowed within " + "`GetSystemPerfData' blocks.", item->key); } } -} -static host_config_t *build_host_config(const oconfig_item_t *ci, const host_config_t *default_host, const service_config_t *def_def_service) { + service->next = host->services; + host->services = service; + + return (0); +} /* }}} int cna_config_system */ + +/* Corresponds to a block. */ +static host_config_t *cna_config_host (const oconfig_item_t *ci, /* {{{ */ + const host_config_t *default_host, const cfg_service_t *def_def_service) +{ int i; oconfig_item_t *item; host_config_t *host, *hc, temp = *default_host; - service_config_t default_service = *def_def_service; + cfg_service_t default_service = *def_def_service; if ((ci->values_num != 1) || (ci->values[0].type != OCONFIG_TYPE_STRING)) { WARNING("netapp plugin: \"Host\" needs exactly one string argument. Ignoring host block."); @@ -1037,7 +1325,7 @@ static host_config_t *build_host_config(const oconfig_item_t *ci, const host_con for (i = 0; i < ci->children_num; ++i) { item = ci->children + i; -// if (!item || !item->key || !*item->key) continue; + /* if (!item || !item->key || !*item->key) continue; */ if (!strcasecmp(item->key, "Address")) { if ((item->values_num != 1) || (item->values[0].type != OCONFIG_TYPE_STRING)) { WARNING("netapp plugin: \"Name\" needs exactly one string argument. Ignoring host block \"%s\".", ci->values[0].value.string); @@ -1071,22 +1359,18 @@ static host_config_t *build_host_config(const oconfig_item_t *ci, const host_con } temp.interval = item->values[0].value.number; } else if (!strcasecmp(item->key, "GetVolumePerfData")) { - build_perf_vol_config(&temp, item); + cna_config_volume_performance(&temp, item); } else if (!strcasecmp(item->key, "GetSystemPerfData")) { - build_perf_sys_config(&temp, item, &default_service); -/* if ((item->values_num != 1) || (item->values[0].type != OCONFIG_TYPE_STRING)) { - WARNING("netapp plugin: \"Collect\" needs exactly one string argument. Ignoring collect block for \"%s\".", ci->values[0].value.string); - continue; - } - build_collect_config(&temp, item);*/ + cna_config_system(&temp, item, &default_service); } else if (!strcasecmp(item->key, "GetWaflPerfData")) { - build_perf_wafl_config(&temp, item); + cna_config_wafl(&temp, item); } else if (!strcasecmp(item->key, "GetDiskPerfData")) { - build_perf_disk_config(&temp, item); + cna_config_disk(&temp, item); } else if (!strcasecmp(item->key, "GetVolumeData")) { - build_volume_config(&temp, item); + cna_config_volume_usage(&temp, item); } else { - WARNING("netapp plugin: Ignoring unknown config option \"%s\" in host block \"%s\".", item->key, ci->values[0].value.string); + WARNING("netapp plugin: Ignoring unknown config option \"%s\" in host block \"%s\".", + item->key, ci->values[0].value.string); } } @@ -1109,31 +1393,114 @@ static host_config_t *build_host_config(const oconfig_item_t *ci, const host_con host->next = host_config; host_config = host; return host; -} +} /* }}} host_config_t *cna_config_host */ -static int build_config (oconfig_item_t *ci) { +/* + * Callbacks registered with the daemon + * + * Pretty standard stuff here. + */ +static int cna_init(void) { /* {{{ */ + char err[256]; + na_elem_t *e; + host_config_t *host; + cfg_service_t *service; + + if (!host_config) { + WARNING("netapp plugin: Plugin loaded but no hosts defined."); + return 1; + } + + if (!na_startup(err, sizeof(err))) { + ERROR("netapp plugin: Error initializing netapp API: %s", err); + return 1; + } + + for (host = host_config; host; host = host->next) { + host->srv = na_server_open(host->host, 1, 1); + na_server_set_transport_type(host->srv, host->protocol, 0); + na_server_set_port(host->srv, host->port); + na_server_style(host->srv, NA_STYLE_LOGIN_PASSWORD); + na_server_adminuser(host->srv, host->username, host->password); + na_server_set_timeout(host->srv, 5); + for (service = host->services; service; service = service->next) { + service->interval = host->interval * service->multiplier; + if (service->handler == collect_perf_system_data) { + service->query = na_elem_new("perf-object-get-instances"); + na_child_add_string(service->query, "objectname", "system"); + } else if (service->handler == query_volume_perf_data) { + service->query = na_elem_new("perf-object-get-instances"); + na_child_add_string(service->query, "objectname", "volume"); +/* e = na_elem_new("instances"); + na_child_add_string(e, "foo", "system"); + na_child_add(root, e);*/ + e = na_elem_new("counters"); + na_child_add_string(e, "foo", "read_ops"); + na_child_add_string(e, "foo", "write_ops"); + na_child_add_string(e, "foo", "read_data"); + na_child_add_string(e, "foo", "write_data"); + na_child_add_string(e, "foo", "read_latency"); + na_child_add_string(e, "foo", "write_latency"); + na_child_add(service->query, e); + } else if (service->handler == query_wafl_data) { + service->query = na_elem_new("perf-object-get-instances"); + na_child_add_string(service->query, "objectname", "wafl"); +/* e = na_elem_new("instances"); + na_child_add_string(e, "foo", "system"); + na_child_add(root, e);*/ + e = na_elem_new("counters"); + na_child_add_string(e, "foo", "name_cache_hit"); + na_child_add_string(e, "foo", "name_cache_miss"); + na_child_add_string(e, "foo", "find_dir_hit"); + na_child_add_string(e, "foo", "find_dir_miss"); + na_child_add_string(e, "foo", "buf_hash_hit"); + na_child_add_string(e, "foo", "buf_hash_miss"); + na_child_add_string(e, "foo", "inode_cache_hit"); + na_child_add_string(e, "foo", "inode_cache_miss"); + /* na_child_add_string(e, "foo", "inode_eject_time"); */ + /* na_child_add_string(e, "foo", "buf_eject_time"); */ + na_child_add(service->query, e); + } else if (service->handler == query_submit_disk_data) { + service->query = na_elem_new("perf-object-get-instances"); + na_child_add_string(service->query, "objectname", "disk"); + e = na_elem_new("counters"); + na_child_add_string(e, "foo", "disk_busy"); + na_child_add_string(e, "foo", "base_for_disk_busy"); + na_child_add(service->query, e); + } else if (service->handler == collect_volume_data) { + service->query = na_elem_new("volume-list-info"); + /* na_child_add_string(service->query, "objectname", "volume"); */ + /* } else if (service->handler == collect_snapshot_data) { */ + /* service->query = na_elem_new("snapshot-list-info"); */ + } + } + } + return 0; +} /* }}} int cna_init */ + +static int cna_config (oconfig_item_t *ci) { /* {{{ */ int i; oconfig_item_t *item; host_config_t default_host = HOST_INIT; - service_config_t default_service = SERVICE_INIT; + cfg_service_t default_service = SERVICE_INIT; for (i = 0; i < ci->children_num; ++i) { item = ci->children + i; -// if (!item || !item->key || !*item->key) continue; + /* if (!item || !item->key || !*item->key) continue; */ if (!strcasecmp(item->key, "Host")) { - build_host_config(item, &default_host, &default_service); + cna_config_host(item, &default_host, &default_service); } else { WARNING("netapp plugin: Ignoring unknown config option \"%s\".", item->key); } } return 0; -} +} /* }}} int cna_config */ -static int netapp_read() { +static int cna_read(void) { /* {{{ */ na_elem_t *out; host_config_t *host; - service_config_t *service; + cfg_service_t *service; for (host = host_config; host; host = host->next) { for (service = host->services; service; service = service->next) { @@ -1145,7 +1512,7 @@ static int netapp_read() { ERROR("netapp plugin: Error %d from host %s: %s", netapp_errno, host->name, na_results_reason(out)); na_elem_free(out); if (netapp_errno == EIO || netapp_errno == ETIMEDOUT) { - // Network problems. Just give up on all other services on this host. + /* Network problems. Just give up on all other services on this host. */ break; } continue; @@ -1155,10 +1522,12 @@ static int netapp_read() { } } return 0; -} +} /* }}} int cna_read */ -void module_register() { - plugin_register_complex_config("netapp", build_config); - plugin_register_init("netapp", config_init); - plugin_register_read("netapp", netapp_read); +void module_register(void) { + plugin_register_complex_config("netapp", cna_config); + plugin_register_init("netapp", cna_init); + plugin_register_read("netapp", cna_read); } + +/* vim: set sw=2 ts=2 noet fdm=marker : */