Velocity • Durability

Version 12.0.0

4. Sample Dashboards and Their Metrics

VDURA provides eight sample dashboards. They are examples for you to import into your own Grafana instance, not part of the VDURA system. To get the dashboard files, request them from your VDURA representative or contact VDURA Technical Support. Their queries show which metrics VDURA treats as primary signals, and they double as worked examples for the tables above.

4.1 PanFS Alerts (panfs-alerts)

vmalert view: Firing counts by severity, category, and node, a firing over time chart, and a table of the ALERTS series.

Panel Type Query
Firing (total) stat count(ALERTS{alertstate=“firing”}) or vector(0)
Critical stat count(ALERTS{alertstate=“firing”,severity=“Critical”}) or vector(0)
Warning stat count(ALERTS{alertstate=“firing”,severity=“Warning”}) or vector(0)
Informational stat count(ALERTS{alertstate=“firing”,severity=“Informational”}) or vector(0)
Firing alerts by severity over time timeseries count by (severity) (ALERTS{alertstate=“firing”})
Currently firing table ALERTS{alertstate=“firing”}
Firing by category bargauge count by (category) (ALERTS{alertstate=“firing”})
Firing by node bargauge count by (node_name) (ALERTS{alertstate=“firing”})

4.2 PanFS API - Audit (panfs-api-audit)

VictoriaLogs only. Reads /var/log/panapi-audit.log with tenant_id and user_id filters. Panels cover events over time by result, failed operations, top action x resource pairs, failures by status_code.

Panel Type Query
Audit events over time (by result) timeseries filename:=“/var/log/panapi-audit.log” tenant_id:$tenant_id user_id:$user_id | stats by (_time:$__interval, result) count() events
Failed operations (selected range) stat filename:=“/var/log/panapi-audit.log” result:=failed tenant_id:$tenant_id user_id:$user_id | stats count() failures
Top operations (action x resource) table filename:=“/var/log/panapi-audit.log” tenant_id:$tenant_id user_id:$user_id | stats by (action, resource) count() requests | sort by (requests desc) limit 20
Failures by status_code table filename:=“/var/log/panapi-audit.log” result:=failed tenant_id:$tenant_id user_id:$user_id | stats by (status_code, action, resource) count() failures | sort by (failures desc) limit 50
Audit log stream logs filename:=“/var/log/panapi-audit.log” tenant_id:$tenant_id user_id:$user_id

4.3 PanFS Logs - Browser (panfs-logs-browser)

VictoriaLogs only. Generic browser over any log file (filename variable) or systemd unit (unit variable) with error/warning/info counts and top collapsed message patterns.

Panel Type Query
Total lines stat filename:$filepath | stats count() total
Errors stat filename:$filepath (level:~“(?i)error|fatal” OR (NOT level:* AND _msg:~“(?i)error|fail”)) | stats count() errors
Warnings stat filename:$filepath (level:~“(?i)warn” OR (NOT level:* AND _msg:~“(?i)warn”)) | stats count() warnings
Info stat filename:$filepath (level:~“(?i)info” OR (NOT level:* AND _msg:~“(?i)info”)) | stats count() info
Top 10 patterns (numbers collapsed) table filename:$filepath | collapse_nums | stats by (_msg) count() hits | sort by (hits desc) limit 10
Errors by file table filename:$filepath (level:~“(?i)error|fatal” OR (NOT level:* AND _msg:~“(?i)error|fail”)) | stats by (filename) count() errors | sort by (errors desc) limit 50
File logs (filename=~$filepath) logs filename:$filepath
Total lines stat unit:$unit | stats count() total
Errors stat unit:$unit (level:~“(?i)error|fatal” OR (NOT level:* AND _msg:~“(?i)error|fail”)) | stats count() errors
Warnings stat unit:$unit (level:~“(?i)warn” OR (NOT level:* AND _msg:~“(?i)warn”)) | stats count() warnings
Info stat unit:$unit (level:~“(?i)info” OR (NOT level:* AND _msg:~“(?i)info”)) | stats count() info
Top 10 patterns (numbers collapsed) table unit:$unit | collapse_nums | stats by (_msg) count() hits | sort by (hits desc) limit 10
Errors by unit table unit:$unit (level:~“(?i)error|fatal” OR (NOT level:* AND _msg:~“(?i)error|fail”)) | stats by (unit) count() errors | sort by (errors desc) limit 50
Journal/unit logs (unit=~$unit) logs unit:$unit

4.4 PanFS Realm Node (panfs-realm-node)

Per node drill down driven by a $Node variable built from label_values(system_uptime, host). It shows CPU per core, top processes by CPU and RSS, memory breakdown, disk I/O and IOPS per device.

Panel Type Query
Per-host CPU utilization (Total) gauge 100 - avg by (host) (cpu_usage_idle{cpu=“cpu-total”, host=~“$Node”})
Per-host CPU utilization (Total) timeseries 100 - avg by (host) (cpu_usage_idle{cpu=“cpu-total”, host=~“$Node”})
CPU utilization (Cores) bargauge 100 - cpu_usage_idle{cpu!=“cpu-total”, host=“$Node”}
Top 10 Processes (CPU Usage) bargauge topk(10, procstat_cpu_usage{host=~“$Node”})
Memory Used gauge mem_used_percent{host=~“$Node”}
Memory Usage timeseries

mem_total{host=~“$Node”}

mem_used{host=~“$Node”}

Memory Usage timeseries

mem_wired{host=~“$Node”}

mem_active{host=~“$Node”}

mem_inactive{host=~“$Node”}

mem_laundry{host=~“$Node”}

mem_free{host=~“$Node”}

Top 10 Processes (Memory Usage) bargauge topk(10, procstat_memory_rss{host=~“$Node”})
Disk IO (Read/Write) timeseries

diskio_read_bytes{host=~“$Node”, name!=“pass0”}

diskio_write_bytes{host=~“$Node”, name!=“pass0”}

Disk IO (Rate[$__rate_interval]) timeseries

rate(diskio_read_bytes{host=~“$Node”, name!=“pass0”}[$__rate_interval])

rate(diskio_write_bytes{host=~“$Node”, name!=“pass0”}[$__rate_interval])

Disk Throughput timeseries sum by (host, node_type) (rate(diskio_read_bytes{host=~“$Node”}[$__rate_interval]) + rate(diskio_write_bytes{host=~“$Node”}[$__rate_interval]))
Total IOPS (read + write) per host timeseries sum by (node_type) (rate(diskio_reads{host=~“$Node”}[$__rate_interval]) + rate(diskio_writes{host=~“$Node”}[$__rate_interval]))

4.5 PanFS Realm Overview (panfs-overall-cluster)

Realm wide view: Node inventory (panfs_node), StorageSet and volume tables, firing alerts by severity, CPU and memory utilization, disk and network throughput by node_type, metadata and free capacity from the iopath fs0..2 gauges, DirectFlow bytes in/out, and log volume from VictoriaLogs. The DirectFlow panels read the iopath measure_df_* metrics, which are last-value gauges (section 3.8), so they show the most recent value and not a sum over the interval.

Panel Type Query
Uptime Cluster Nodes table count by (node_type) (system_uptime)
Storage table

count(group by (name) (panfs_storageset > 0))

count(group by (volume_id) (panfs_volume > 0))

Firing Alerts bargauge count by (severity) (ALERTS{alertstate=“firing”})
Versions table group by (software_revision) (last_over_time(panfs_node[$__rate_interval]))
CPU Utilization stat 100 - avg by (host) (cpu_usage_idle{cpu=“cpu-total”})
Memory Utilization stat mem_used_percent
Cluster Nodes table group by (name, ip_address, node_type, node_name, node_state, node_data_serial, node_hardware_serial, node_state_name, node_state_description, enclosure_name, software_revision, cpu_model, memory_bytes, cpu_active_cores, internal_temperature_celsius, cpu_temperature_celsius) (last_over_time(panfs_node[$__rate_interval]))
Storage Sets table topk(1, last_over_time(panfs_storageset[5m]>0)) by (name)
Volumes table topk(1, last_over_time(panfs_volume[5m])) by (volume_id)
Active alert messages state-timeline

max by (message) (

(ALERTS{severity=“Critical”}) * 3

or (ALERTS{severity=“Warning”}) * 2

or (ALERTS{severity=“Informational”}) * 1

)

StorageSets Alerts table ALERTS{message=~“.(BladeSet |Storage[ ]?[Ss]et ).”}
Volume Alerts table ALERTS{message=~“.(Volume ).”}
Software Alerts table ALERTS{message=~“(Software Process.|The process .restarted)”}
Disk Throughput timeseries sum by (host, node_type) (rate(diskio_read_bytes[$__rate_interval]) + rate(diskio_write_bytes[$__rate_interval]))
Total IOPS (read + write) per host timeseries sum by (host, node_type) (rate(diskio_reads[$__rate_interval]) + rate(diskio_writes[$__rate_interval]))
CPU Utilization timeseries avg by (host) (100 - cpu_usage_idle{cpu=“cpu-total”})
Disk IO Usage timeseries

(sum(rate(diskio_read_bytes{node_type=“vpod”}[$__rate_interval])) + sum(rate(diskio_write_bytes{node_type=“vpod”}[$__rate_interval])))

sum(rate(diskio_reads{node_type=“vpod”}[$__rate_interval])) + sum(rate(diskio_writes{node_type=“vpod”}[$__rate_interval]))

Throughput Out timeseries sum(rate(net_bytes_sent{interface=~“eth0|bond0”}[$__rate_interval]) * 8) by (node_type)
Throughput In timeseries sum(rate(net_bytes_recv[$__rate_interval]) * 8) by (node_type)
Metadata Used timeseries sum(magent_local_sw_iopath_fs0_metadata_capacity+magent_local_sw_iopath_fs1_metadata_capacity+magent_local_sw_iopath_fs2_metadata_capacity)
Total Capacity (Free) timeseries sum(last_over_time(magent_local_sw_iopath_fs0_free_capacity[$__rate_interval]))
IOPath Measure (DF Bytes In/Out) timeseries

sum(last_over_time(magent_local_sw_iopath_measure_df_bytes_in[$__rate_interval]))

sum(last_over_time(magent_local_sw_iopath_measure_df_bytes_out[$__rate_interval]))

Duration timeseries sum(last_over_time(magent_local_sw_iopath_measure_df_io_duration_usec[$__rate_interval]))
Log volume (all logs) timeseries
  • stats by (_time:$__interval) count() logs
Log rate (total vs error), per second timeseries

  • stats by (_time:$__interval) rate() log_rate

    i(error) | stats by (_time:$__interval) rate() error_rate

4.6 VersityGW S3 Metrics (panfs-s3)

VictoriaMetrics only. Request, error, throughput, and object counters for the S3 gateways of one tenant (metrics in section 3.9). Two variables narrow the view, Tenant (the values of the service label of versitygw_failed_count_value) and Instance (service), a multi-select list of the services whose names start with the chosen tenant followed by two underscores. The default time range is the last hour, refreshed every 30 seconds.

Panel Type Query
Failed req/s stat sum(rate(versitygw_failed_count_value{service=~“$service”}[$__rate_interval]))
Success req/s stat sum(rate(versitygw_success_count_value{service=~“$service”}[$__rate_interval]))
Error ratio stat 100 * sum(rate(versitygw_failed_count_value{service=~“$service”}[$__rate_interval])) / clamp_min(sum(rate(versitygw_failed_count_value{service=~“$service”}[$__rate_interval])) + sum(rate(versitygw_success_count_value{service=~“$service”}[$__rate_interval])), 0.0001)
Throughput (read+write) stat sum(rate(versitygw_bytes_read_value{service=~“$service”}[$__rate_interval])) + sum(rate(versitygw_bytes_written_value{service=~“$service”}[$__rate_interval]))
Request rate — success vs failed timeseries sum(rate(versitygw_success_count_value{service=~“$service”}[$__rate_interval])) (success) and sum(rate(versitygw_failed_count_value{service=~“$service”}[$__rate_interval])) (failed)
Failed rate by instance timeseries sum by (service) (rate(versitygw_failed_count_value{service=~“$service”}[$__rate_interval]))
Rate by action timeseries sum by (action) (rate(versitygw_failed_count_value{service=~“$service”}[$__rate_interval]))
Rate by HTTP status timeseries sum by (status) (rate(versitygw_failed_count_value{service=~“$service”}[$__rate_interval]))
Throughput — bytes/s timeseries sum(rate(versitygw_bytes_read_value{service=~“$service”}[$__rate_interval])) (read) and sum(rate(versitygw_bytes_written_value{service=~“$service”}[$__rate_interval])) (written)
Objects/s — created vs removed timeseries sum(rate(versitygw_object_created_count_value{service=~“$service”}[$__rate_interval])) (created) and sum(rate(versitygw_object_removed_count_value{service=~“$service”}[$__rate_interval])) (removed)

4.7 PanFS Hardware Health (panfs-hardware)

VictoriaMetrics only. Hardware health from the management agent on each node (the magent_local_hw_* metrics and related agent metrics) and from the realm database (the panfs_* metrics). Panels cover hardware status, temperatures and thermal alarms, NICs, memory, CPU and system disks, core files, power supplies, fans, switches, drive inventory and wear, and the status of the realm collectors. The temperature panels read a flat 0 where there are no sensors. A chassis whose firmware does not report power supplies or fans also leaves those two tables empty, and the Switches table is expected to be empty on current hardware. Treat an empty or flat-zero panel as missing data, not as proof of health. The hardware status value is 0 for OK and 1 for FAILED (section 3.4), so the “Nodes with non-OK hardware status” panel, which tests for a value other than 1, counts the nodes that are not FAILED. The NIC slot status panel always shows 0 because the agent never changes the NIC status; the NIC link warnings panel carries the NIC condition. The fan-failure trip and drives over temp series in the Thermal alarms panel are never set and always read 0. The System disk utilisation panel shows the highest percent busy of the local drives, not how full the system disk is. The dashboard has no variables; the default time range is the last hour, refreshed every 30 seconds.

Panel Type Query
Nodes reporting hardware status stat count(magent_local_hw_status)
Nodes with non-OK hardware status stat count(magent_local_hw_status != 1) or vector(0)
Hardware status by node bargauge magent_local_hw_status
Enclosure temperatures timeseries magent_local_hw_internal_temp ({{node_name}} internal)
magent_local_hw_external_temp ({{node_name}} external)
magent_local_hw_incoming_air_temp ({{node_name}} incoming air)
CPU and node temperatures (realm inventory) timeseries label_value(last_over_time(panfs_node[5m]), “cpu_temperature_celsius”) ({{node_name}} cpu)
label_value(last_over_time(panfs_node[5m]), “internal_temperature_celsius”) ({{node_name}} internal)
Thermal alarms timeseries sum(magent_local_hw_high_temp_alert) (high temp)
sum(magent_local_hw_trip_temp_alert) (trip temp)
sum(magent_local_hw_ff_trip_temp_alert) (fan-failure trip)
sum(magent_local_hw_temperature_shutdown) (thermal shutdown)
sum(magent_local_hw_drives_over_temp) (drives over temp)
Nodes in thermal alarm bargauge magent_local_hw_high_temp_alert + magent_local_hw_trip_temp_alert + magent_local_hw_temperature_shutdown
NIC slot status timeseries sum(magent_local_hw_nic1_status) (nic1 status)
sum(magent_local_hw_nic2_status) (nic2 status)
sum(magent_local_hw_nic3_status) (nic3 status)
sum(magent_local_hw_nic4_status) (nic4 status)
NIC link warnings timeseries sum(magent_local_hw_nic1_nic_warnings) (nic1 warnings)
sum(magent_local_hw_nic2_nic_warnings) (nic2 warnings)
sum(magent_local_hw_nic3_nic_warnings) (nic3 warnings)
sum(magent_local_hw_nic4_nic_warnings) (nic4 warnings)
NICs with uplink down stat count(last_over_time(panfs_nic{uplink_down=“true”}[5m])) or vector(0)
NIC inventory table group by (node_name, interface, device, nic_type, active, in_use, uplink_down, mode_of_operation, firmware, warnings_disabled) (last_over_time(panfs_nic[$__rate_interval]))
IPMI sensors table group by (node_name, node_ip_address, ipmi_ip_address, mac_address, firmware_revision, netmask, default_router) (last_over_time(panfs_ipmi[$__rate_interval]))
Installed memory per node bargauge magent_local_hw_mem * 1048576
CPU cores per node bargauge magent_local_hw_cpu_active_cores ({{node_name}} active)
magent_local_hw_cpu_hw_threads ({{node_name}} threads)
System disk utilisation bargauge magent_local_sw_pm_disk_util
System disk capacity used timeseries magent_local_sw_pm_cpty_used ({{node_name}} used)
magent_local_sw_pm_cpty_size ({{node_name}} total)
Core files recorded timeseries count by (host) (magent_local_sw_monitor_corefile)
Power supplies table last_over_time(panfs_hw_power_supply[$__rate_interval])
Fans table last_over_time(panfs_hw_fan[$__rate_interval])
Switches [blocked on PanFS platform fix] table group by (enclosure_name, switch_number, firmware_revision, kernel_revision, mac_address, flooding) (last_over_time(panfs_switch[$__rate_interval]))
Drives table group by (node_name, drive_number, name, type, vendor, model, hardware_serial, firmware_version, hardware_status, ssd_remaining_lifetime, reallocation_count, lun) (last_over_time(panfs_drive[$__rate_interval]))
Drive throughput timeseries sum by (node_name) (rate(panfs_drive_bytes_read[$__rate_interval])) ({{node_name}} read)
sum by (node_name) (rate(panfs_drive_bytes_written[$__rate_interval])) ({{node_name}} write)
Drive status by slot timeseries label_replace(sum by (__name__) ({__name__=~“magent_local_hw_hd[0-9]+_status”}), “slot”, “$1”, “__name__”, “magent_local_hw_(hd[0-9]+)_status”)
SSD wearout indicator bargauge {__name__=~“magent_local_hw_hd[0-9]+_ssd_wearout_indicator_value”}
Drive reallocation count timeseries {__name__=~“magent_local_hw_hd[0-9]+_hdd_reallocation_count”}
Drive reserve-space and wearout alerts timeseries sum({__name__=~“magent_local_hw_hd[0-9]+_ssd_wearout_indicator_alert”}) (wearout alerts)
sum({__name__=~“magent_local_hw_hd[0-9]+_ssd_available_reserve_space_alert”}) (reserve-space alerts)
Realm collectors failing stat count({__name__=~“panfs_.+_scrape_success”} == 0) or vector(0)
Realm collector scrape status bargauge label_replace({__name__=~“panfs_.+_scrape_success”}, “collector”, “$1”, “__name__”, “panfs_(.+)_scrape_success”)

4.8 PanFS Performance (panfs-performance)

VictoriaMetrics only. File Manager (metadata) operation rates and latency, the metadata caches, OSD I/O (throughput, operations, latency breakdown, task queue depth, errors and timeouts, iSCSI connections), raw capacity and free space, and volume and StorageSet health counters. It reads the magent_local_sw_fm_*, magent_local_sw_gen_cache_* and magent_local_sw_iopath_* metrics (sections 3.6 to 3.8) and the panfs_volume_* and panfs_storageset_* metrics (section 3.1). The dashboard has no variables; the default time range is the last hour, refreshed every 30 seconds. Several panels do not show what their titles say. The FM and cache latency panels divide a lifetime value by a count, but each lifetime value is the duration of the last operation in nanoseconds and not an accumulated total (sections 3.6 and 3.7). The metadata cache hit ratio panels divide misses by lookups only, while the correct divisor is lookups plus gets. The OSD throughput, OSD IOPS, and mean OSD I/O latency panels sum the iopath measure_* gauges, which hold the last value and flags (section 3.8), and divide by 20 seconds, so they do not show a rate. The per-command counts and duration series are the cumulative source. The Realm raw capacity used, Realm raw capacity free, and Free capacity per storage node panels read remaining_capacity_bytes, which includes the spare bucket (section 3.8), so they count reserved spare space as free; free space for new data is the bucket0 remaining value.

Panel Type Query
Metadata operation rate (all FM ops) timeseries sum by (host) (rate(magent_local_sw_fm_global_op_count[$__rate_interval]))
Mean metadata latency (all FM ops) timeseries sum by (host) (rate(magent_local_sw_fm_global_op_lifetime[$__rate_interval])) / sum by (host) (rate(magent_local_sw_fm_global_op_count[$__rate_interval]))
Top 10 metadata operations by rate bargauge topk(10, sum by (op) (label_replace(rate({__name__=~“magent_local_sw_fm_op_count_.+”}[$__rate_interval]) keep_metric_names, “op”, “$1”, “__name__”, “magent_local_sw_fm_op_count_(.+)”)))
Mean latency by metadata operation (top 10) bargauge topk(10, sum by (op) (label_replace(rate({__name__=~“magent_local_sw_fm_op_lifetime_.+”}[$__rate_interval]) keep_metric_names, “op”, “$1”, “__name__”, “magent_local_sw_fm_op_lifetime_(.+)”)))
Metadata cache hit ratio timeseries 1 - sum(rate(magent_local_sw_gen_cache_cache_FM_dir_cache_miss_count[$__rate_interval])) / sum(rate(magent_local_sw_gen_cache_cache_FM_dir_cache_lookup_count[$__rate_interval])) (FM_dir_cache)
1 - sum(rate(magent_local_sw_gen_cache_cache_FM_object_attribute_cache_miss_count[$__rate_interval])) / sum(rate(magent_local_sw_gen_cache_cache_FM_object_attribute_cache_lookup_count[$__rate_interval])) (FM_object_attribute_cache)
1 - sum(rate(magent_local_sw_gen_cache_cache_FM_ogroup_cache_miss_count[$__rate_interval])) / sum(rate(magent_local_sw_gen_cache_cache_FM_ogroup_cache_lookup_count[$__rate_interval])) (FM_ogroup_cache)
1 - sum(rate(magent_local_sw_gen_cache_cache_SM_object_cache_miss_count[$__rate_interval])) / sum(rate(magent_local_sw_gen_cache_cache_SM_object_cache_lookup_count[$__rate_interval])) (SM_object_cache)
1 - sum(rate(magent_local_sw_gen_cache_cache_pan_id_user_info_cache_miss_count[$__rate_interval])) / sum(rate(magent_local_sw_gen_cache_cache_pan_id_user_info_cache_lookup_count[$__rate_interval])) (pan_id_user_info_cache)
Cache partition pressure (used / limit) bargauge sum(magent_local_sw_gen_cache_partition_fm_dir_bytes_cur) / sum(magent_local_sw_gen_cache_partition_fm_dir_bytes_limit) (fm_dir)
sum(magent_local_sw_gen_cache_partition_fm_attr_bytes_cur) / sum(magent_local_sw_gen_cache_partition_fm_attr_bytes_limit) (fm_attr)
sum(magent_local_sw_gen_cache_partition_fm_ogroup_bytes_cur) / sum(magent_local_sw_gen_cache_partition_fm_ogroup_bytes_limit) (fm_ogroup)
sum(magent_local_sw_gen_cache_partition_sm_obj_bytes_cur) / sum(magent_local_sw_gen_cache_partition_sm_obj_bytes_limit) (sm_obj)
sum(magent_local_sw_gen_cache_partition_default_bytes_cur) / sum(magent_local_sw_gen_cache_partition_default_bytes_limit) (default)
Cache evictions timeseries sum(rate(magent_local_sw_gen_cache_partition_fm_dir_evicted_ent_count[$__rate_interval])) (fm_dir)
sum(rate(magent_local_sw_gen_cache_partition_fm_attr_evicted_ent_count[$__rate_interval])) (fm_attr)
sum(rate(magent_local_sw_gen_cache_partition_fm_ogroup_evicted_ent_count[$__rate_interval])) (fm_ogroup)
sum(rate(magent_local_sw_gen_cache_partition_sm_obj_evicted_ent_count[$__rate_interval])) (sm_obj)
sum(rate(magent_local_sw_gen_cache_partition_default_evicted_ent_count[$__rate_interval])) (default)
Cache backing-store I/O latency timeseries sum(rate(magent_local_sw_gen_cache_cache_FM_dir_cache_load_io_lifetime[$__rate_interval])) / sum(rate(magent_local_sw_gen_cache_cache_FM_dir_cache_load_io_count[$__rate_interval])) (FM_dir_cache)
sum(rate(magent_local_sw_gen_cache_cache_FM_object_attribute_cache_load_io_lifetime[$__rate_interval])) / sum(rate(magent_local_sw_gen_cache_cache_FM_object_attribute_cache_load_io_count[$__rate_interval])) (FM_object_attribute_cache)
sum(rate(magent_local_sw_gen_cache_cache_FM_ogroup_cache_load_io_lifetime[$__rate_interval])) / sum(rate(magent_local_sw_gen_cache_cache_FM_ogroup_cache_load_io_count[$__rate_interval])) (FM_ogroup_cache)
sum(rate(magent_local_sw_gen_cache_cache_SM_object_cache_load_io_lifetime[$__rate_interval])) / sum(rate(magent_local_sw_gen_cache_cache_SM_object_cache_load_io_count[$__rate_interval])) (SM_object_cache)
sum(rate(magent_local_sw_gen_cache_cache_pan_id_user_info_cache_load_io_lifetime[$__rate_interval])) / sum(rate(magent_local_sw_gen_cache_cache_pan_id_user_info_cache_load_io_count[$__rate_interval])) (pan_id_user_info_cache)
OSD throughput timeseries sum(magent_local_sw_iopath_measure_bytes_in) / 20 (in)
sum(magent_local_sw_iopath_measure_bytes_out) / 20 (out)
OSD IOPS timeseries sum(magent_local_sw_iopath_measure_io_occured) / 20 (total)
sum(magent_local_sw_iopath_measure_op_count_write) / 20 (write)
sum(magent_local_sw_iopath_measure_op_count_non_write) / 20 (non-write)
OSD latency breakdown (queue vs osdfs vs wire) timeseries sum({__name__=~“magent_local_sw_iopath_durationall_.+”}) (all)
sum({__name__=~“magent_local_sw_iopath_durationqueue_.+”}) (queue)
sum({__name__=~“magent_local_sw_iopath_durationosdfs_.+”}) (osdfs)
sum({__name__=~“magent_local_sw_iopath_durationwire_.+”}) (wire)
sum({__name__=~“magent_local_sw_iopath_durationactive_.+”}) (active)
Mean OSD I/O latency by class timeseries sum(magent_local_sw_iopath_measure_io_duration_usec) / sum(magent_local_sw_iopath_measure_io_occured) (all)
sum(magent_local_sw_iopath_measure_rd_duration_usec) / sum(magent_local_sw_iopath_measure_io_occured) (read)
sum(magent_local_sw_iopath_measure_wr_duration_usec) / sum(magent_local_sw_iopath_measure_io_occured) (write)
sum(magent_local_sw_iopath_measure_ra_duration_usec) / sum(magent_local_sw_iopath_measure_io_occured) (readahead)
OSD task queue depth timeseries sum by (host) (magent_local_sw_iopath_measure_task_queue_depth)
OSD errors and timeouts timeseries sum(increase(magent_local_sw_iopath_measure_num_timeout_cmds[$__rate_interval])) (command timeouts)
sum(increase(magent_local_sw_iopath_measure_num_timeout_sessions[$__rate_interval])) (session timeouts)
sum(increase(magent_local_sw_iopath_measure_num_failed_allocations[$__rate_interval])) (failed allocations)
iSCSI connections timeseries sum by (host) (magent_local_sw_iopath_iscsi_num_cur_conns)
Realm raw capacity used gauge 1 - sum(magent_local_sw_iopath_measure_remaining_capacity_bytes) / sum(magent_local_sw_iopath_measure_capacity_bytes)
Realm raw capacity free stat sum(magent_local_sw_iopath_measure_remaining_capacity_bytes)
Free capacity per storage node bargauge sum by (host) (magent_local_sw_iopath_measure_remaining_capacity_bytes)
Capacity by bucket timeseries sum(magent_local_sw_iopath_measure_bucket0_capacity_bytes) (bucket0 total)
sum(magent_local_sw_iopath_measure_bucket0_remaining_capacity_bytes) (bucket0 free)
sum(magent_local_sw_iopath_measure_bucket1_capacity_bytes) (bucket1 total)
sum(magent_local_sw_iopath_measure_bucket1_remaining_capacity_bytes) (bucket1 free)
OSD local filesystem capacity timeseries sum(magent_local_sw_iopath_fs0_data_capacity + magent_local_sw_iopath_fs1_data_capacity + magent_local_sw_iopath_fs2_data_capacity) (data)
sum(magent_local_sw_iopath_fs0_metadata_capacity + magent_local_sw_iopath_fs1_metadata_capacity + magent_local_sw_iopath_fs2_metadata_capacity) (metadata)
sum(magent_local_sw_iopath_fs0_free_capacity) (free (fs0))
Volume space used timeseries panfs_volume_space_used
Volume quota consumption bargauge panfs_volume_quota_consumed_bytes / panfs_volume_hard_quota_bytes
Volume integrity counters timeseries sum(panfs_volume_corrupt_object_count) (corrupt objects)
sum(panfs_volume_corrupt_parity_object_count) (corrupt parity objects)
sum(panfs_volume_corrupt_log_count) (corrupt logs)
sum(panfs_volume_down_storage_node_count) (down storage nodes)
StorageSet imbalance and errors timeseries panfs_storageset_imbalance_bytes ({{name}} imbalance)
panfs_storageset_blade_imbalance_bytes ({{name}} blade imbalance)
sum(panfs_storageset_uncorrectable_error_count) (uncorrectable errors)
sum(panfs_storageset_correctable_error_count) (correctable errors)