Metrics Reference

All metrics are gauges prefixed powerstore_ and carry an array label (the configured array name) plus cluster_id (the PowerStore system id).

Meta & health metrics

Metric	Labels	Meaning
powerstore_up	array	1 if the array was scraped successfully this cycle, else 0.
powerstore_last_scrape_timestamp_seconds	array	Unix time of the last successful collection.
powerstore_array_bulk_api	array	1 if bulk CSV collection is active for this array, 0 if per-entity REST fallback is used.
pstore_exporter_build_info	version, goversion	Exporter build information; constant 1, exposing the running exporter version and Go version. Exporter-level (no array/cluster_id label).

Volume metrics

Each volume metric carries: array, cluster_id, volume_name, volume_id, appliance_id, appliance_name, volume_group_name, volume_group_id.

Metric	Unit	Description
powerstore_volume_read_iops	ops/s	Read I/O operations per second
powerstore_volume_write_iops	ops/s	Write I/O operations per second
powerstore_volume_total_iops	ops/s	Total (read + write) I/O operations per second
powerstore_volume_read_bandwidth_bytes_per_second	bytes/s	Read throughput
powerstore_volume_write_bandwidth_bytes_per_second	bytes/s	Write throughput
powerstore_volume_read_latency_microseconds	µs	Average read latency
powerstore_volume_write_latency_microseconds	µs	Average write latency
powerstore_volume_avg_io_size_bytes	bytes	Average I/O size (read + write combined)

Appliance metrics

Each appliance metric carries: array, cluster_id, appliance_name, appliance_id, service_tag.

Metric	Unit	Description
powerstore_appliance_read_iops	ops/s	Appliance read I/O operations per second
powerstore_appliance_write_iops	ops/s	Appliance write I/O operations per second
powerstore_appliance_total_iops	ops/s	Appliance total I/O operations per second
powerstore_appliance_read_bandwidth_bytes_per_second	bytes/s	Appliance read throughput
powerstore_appliance_write_bandwidth_bytes_per_second	bytes/s	Appliance write throughput
powerstore_appliance_read_latency_microseconds	µs	Appliance average read latency
powerstore_appliance_write_latency_microseconds	µs	Appliance average write latency
powerstore_appliance_io_workload_cpu_utilization	ratio (0–1)	CPU utilization fraction for I/O workload
powerstore_appliance_physical_total_bytes	bytes	Raw physical capacity
powerstore_appliance_physical_used_bytes	bytes	Physical capacity in use
powerstore_appliance_logical_provisioned_bytes	bytes	Total logically provisioned capacity
powerstore_appliance_logical_used_bytes	bytes	Logical capacity consumed by data
powerstore_appliance_data_reduction_ratio	ratio	Data reduction ratio (compression + dedup)
powerstore_appliance_efficiency_ratio	ratio	Overall storage efficiency ratio
powerstore_appliance_snapshot_savings_ratio	ratio	Savings attributable to snapshots
powerstore_appliance_thin_savings_ratio	ratio	Savings attributable to thin provisioning

Cluster capacity metrics

Cluster-wide space rollup for capacity forecasting (SpaceMetricsByCluster). Carries only array, cluster_id.

Metric	Unit	Description
powerstore_cluster_physical_total_bytes	bytes	Total physical capacity of the cluster
powerstore_cluster_physical_used_bytes	bytes	Physical space consumed
powerstore_cluster_logical_provisioned_bytes	bytes	Total provisioned logical capacity
powerstore_cluster_logical_used_bytes	bytes	Logical space written
powerstore_cluster_data_reduction_ratio	ratio	Cluster data-reduction ratio (N:1)
powerstore_cluster_efficiency_ratio	ratio	Overall cluster efficiency ratio (N:1)

Volume group metrics

Each volume-group metric carries: array, cluster_id, volume_group_name, volume_group_id. Performance comes from PerformanceMetricsByVg.

Metric	Unit	Description
powerstore_volume_group_read_iops	ops/s	Read I/O operations per second
powerstore_volume_group_write_iops	ops/s	Write I/O operations per second
powerstore_volume_group_total_iops	ops/s	Total I/O operations per second
powerstore_volume_group_read_bandwidth_bytes_per_second	bytes/s	Read throughput
powerstore_volume_group_write_bandwidth_bytes_per_second	bytes/s	Write throughput
powerstore_volume_group_read_latency_microseconds	µs	Average read latency
powerstore_volume_group_write_latency_microseconds	µs	Average write latency
powerstore_volume_group_avg_io_size_bytes	bytes	Average I/O size

File system metrics

Each file system metric carries: array, cluster_id, file_system_name, file_system_id, nas_server_name, nas_server_id.

Capacity is inventory-derived; performance comes from the typed PerformanceMetricsByFileSystem method (averaged counters), parallel to the volume performance metrics. Both are emitted on the bulk and per-entity paths.

Metric	Unit	Description
powerstore_file_system_size_total_bytes	bytes	Provisioned file system capacity
powerstore_file_system_size_used_bytes	bytes	Used file system capacity
powerstore_file_system_read_iops	ops/s	Average read I/O operations per second
powerstore_file_system_write_iops	ops/s	Average write I/O operations per second
powerstore_file_system_total_iops	ops/s	Average total I/O operations per second
powerstore_file_system_read_bandwidth_bytes_per_second	bytes/s	Average read throughput
powerstore_file_system_write_bandwidth_bytes_per_second	bytes/s	Average write throughput
powerstore_file_system_read_latency_microseconds	µs	Average read latency
powerstore_file_system_write_latency_microseconds	µs	Average write latency
powerstore_file_system_avg_io_size_bytes	bytes	Average I/O size

Port metrics

Each port metric carries: array, cluster_id, port_name, port_id, port_type (eth or fc), appliance_id.

Metric	Labels	Description
powerstore_port_link_up	(port labels)	1 if the port link is up, 0 if down.

Drive metrics

Each drive metric carries: array, cluster_id, drive_id, drive_name, appliance_id, and is emitted once per drive on the array. Drives are enumerated via a single generic GET on the hardware resource (PowerStore exposes no typed list-drives method; see ADR-0009). Prior to the lifecycle_state property fix, the enumeration query 400'd and both metrics below were silently absent from /metrics.

Metric	Labels	Description
powerstore_drive_state	state	Info series, always 1; the drive's lifecycle state is the state label (Healthy, Failed, …).
powerstore_drive_wear_level_ratio	(drive labels)	SSD wear consumed, 0.0 (new) → 1.0 (worn out). Only emitted on PowerStoreOS ≥ 4.3.0.0, which reports drive_wear_level.

Alert metrics

Each alert metric carries: array, cluster_id, severity.

Metric	Labels	Description
powerstore_alert_active	severity	Count of active (uncleared) alerts at this severity.

Alerts are aggregated by severity (not one series per alert) to keep cardinality bounded. The standard PowerStore severities — Critical, Major, Minor, Info, None — are always emitted, with value 0 when no active alert matches, so alerting rules can rely on a stable series (e.g. powerstore_alert_active{severity="Critical"} > 0) rather than one that disappears when the array is healthy. Source: gopowerstore GetAlerts. Emitted on both the bulk and per-entity paths (see ADR-0009).

Replication metrics

Source: GetReplicationRules (typed) for RPO; replication sessions are enumerated via the generic API (replication_session, the same escape hatch used for drives and the Metro witness — see ADR-0009), which captures every session on the array, including Metro sessions (role=Metro_Preferred / Metro_Non_Preferred, type=Metro_Active_Active) that are not driven by a protection policy and were previously invisible. Transfer/backlog (VolumeMirrorTransferRate, typed) is then queried only for the volume-type sessions found, so a policy-bearing volume with no session no longer emits a phantom 0/0 transfer series. Emitted on both export paths.

Each replication and witness metric carries: array, cluster_id.

Metric	Labels	Description
powerstore_replication_session_state	session_id, local_resource_id, local_resource_name, resource_type, role, type, remote_system_id, state	Info series, always 1, one per replication session of any resource type (e.g. volume, volume_group, file_system) — including Metro sessions. The session's current state is the state label (OK, Synchronizing, Error, Fractured, System_Paused, Paused, …); role/type distinguish Metro (role=Metro_Preferred|Metro_Non_Preferred, type=Metro_Active_Active) from async/sync sessions. local_resource_name is the replicated resource's name, falling back to its id when the resource is absent from the inventory.
powerstore_replication_rpo_seconds	rule_id, remote_system_id	Configured RPO of a replication rule, in seconds (0 = synchronous).
powerstore_replication_transfer_rate_bytes_per_second	resource_id, resource_name, resource_type	Current mirror replication throughput for the resource. Emitted only for volumes with an active replication session.
powerstore_replication_data_remaining_bytes	resource_id, resource_name, resource_type	Outstanding data still to be replicated (backlog / RPO-risk indicator). Emitted only for volumes with an active replication session.
powerstore_metro_witness_state	witness_id, witness_name, state	Info series, always 1; the Metro witness service's overall state is the state label (OK, Partially_Connected, Disconnected, Initializing, Deleting).
powerstore_metro_witness_connection_state	witness_id, appliance_id, node_id, state	Info series, always 1; one per node, with the node's connection to the witness in the state label (OK, Disconnected, Initializing).

All _state info metrics follow the enum/info idiom: alert on undesirable states with a label matcher rather than a numeric threshold — powerstore_replication_session_state, powerstore_metro_witness_state, and powerstore_metro_witness_connection_state all carry the value 1 regardless of state.

PromQL guidance

Do not use rate()

powerstore_* IOPS and bandwidth values are already per-second gauges computed by PowerStore. In PromQL aggregate them with sum / avg by labels — never wrap them in rate() or you will double-rate the values.

# Total write IOPS across all volumes on one array
sum by (array) (powerstore_volume_write_iops{array="pstore-prod"})

# Average appliance read latency
avg by (appliance_name) (powerstore_appliance_read_latency_microseconds)

# File-system capacity used %
100 * powerstore_file_system_size_used_bytes / powerstore_file_system_size_total_bytes

# Physical capacity utilization per appliance
100 * powerstore_appliance_physical_used_bytes / powerstore_appliance_physical_total_bytes

# Cluster physical capacity utilization %
100 * powerstore_cluster_physical_used_bytes / powerstore_cluster_physical_total_bytes

# Ports that are down
powerstore_port_link_up == 0

# Any active critical alerts, per array
sum by (array) (powerstore_alert_active{severity="Critical"}) > 0

# Replication sessions in a bad state
powerstore_replication_session_state{state=~"Error|Fractured|System_Paused|Paused"}

# Replication backlog exceeding 1 GiB
powerstore_replication_data_remaining_bytes > 1073741824

# Metro witness degraded or unreachable
powerstore_metro_witness_state{state=~"Disconnected|Partially_Connected"}
# A specific node cannot reach the witness
powerstore_metro_witness_connection_state{state="Disconnected"}

# Drives that are not healthy
powerstore_drive_state{state!="Healthy"}

# SSDs past 80% wear
powerstore_drive_wear_level_ratio > 0.8

Deferred metrics (available in the API, not yet wired)

The following are exposed by gopowerstore v1.22.0 as typed methods and are available, not unavailable — they are simply not collected yet. Implementation order and field mappings are in reconciliation-2026-06-05.md and ADR-0009.

• Per-volume space (powerstore_volume_logical_used_bytes) — per-volume growth via SpaceMetricsByVolume. Deferred: it adds one API call per volume (doubling the per-volume call volume); cluster-level forecasting via powerstore_cluster_* covers the common case first. (P3)
• NAS/SMB/NFS per-protocol & per-node performance — PerformanceMetricsNfs*ByNode, PerformanceMetricsSmb*ByNode, PerformanceMetricsByNode. Deferred under YAGNI. (P3)