sendnrw/windows_exporter
2
0
Fork 0
mirror of https://github.com/prometheus-community/windows_exporter.git synced 2026-10-20 19:49:03 +02:00
Code Issues Packages Projects Releases Wiki Activity

collector: name instance labels after the entity they identify (#2744)

Browse Source 
Signed-off-by: Jan-Otto Kröpke <mail@jkroepke.de>
This commit is contained in:
Jan-Otto Kröpke
2026-10-16 11:07:04 +02:00
committed by GitHub
parent 3fb18521f6
commit 667200f757
67 changed files with 1960 additions and 751 deletions
Download Patch File Download Diff File Expand all files Collapse all files
UPGRADE.md
+54 -2
View File
@@ -19,6 +19,7 @@ Most installations need no changes. Go through the checklist below and read the
| You alert on BitLocker status | Replace `status="disabled"` with `status="off"` | [BitLocker status](#bitlocker-status) |
| You query Active Directory latency or security descriptor propagation, DFSR, DNS dynamic update, Hyper-V, logical disk queue, cache, IIS, memory, RemoteFX, SQL Server, scheduled task, SMB server share or terminal services metrics | Update queries | [Renamed, removed and retyped metrics](#renamed-removed-and-retyped-metrics) |
| You query metrics in MHz, percent, minutes, milliseconds, microseconds, kilobytes, megabytes or parts per billion, counters without a `_total` suffix, or Exchange, Hyper-V DataStore and Virtual SMB, IIS application pool, mscluster, NPS or SQL Server counters | Update queries | [Base units and counter names](#base-units-and-counter-names) |
| You query `dfsr`, `dhcp` scope, `diskdrive`, `dns` WMI, `exchange`, Hyper-V Virtual SMB, `mscluster`, `msmq`, `registry`, `service`, `storage_spaces` or `wmi` metrics by their `name` or `instance` label, or `performancecounter` objects without `instance_label` | Rename the label in queries and alerts | [Instance label names](#instance-label-names) |
| You enable the `scheduled_task` collector and use tasks below `\Microsoft\` or missed runs | Set `--collector.scheduled_task.exclude` or `missed-runs` | [scheduled_task collector](#scheduled_task-collector) |
| You chart SQL Server latencies, lock counts or replica queues, DHCP denials, AD LSA writes, AD computer creations, AD FS database query times, DNS zone transfers or RemoteFX skipped frames | Expect corrected values | [Corrected values](#corrected-values) |
| You enable the `hyperv` or `mscluster` collector | Check the new default sub-collectors | [New default sub-collectors](#new-default-sub-collectors) |
@@ -86,7 +87,7 @@ collector:
metric: windows_thermalzone_throttle_reasons
```
`instance_label: name` keeps the `name` label of 0.31. Without it, the label is called `instance`, which collides with the target label that Prometheus adds.
`instance_label: name` keeps the `name` label of 0.31. Without it, the label is called `object_instance`. In 0.31 the default label was `instance`, which Prometheus stored as `exported_instance` (or, with `honor_labels: true`, as `instance`, replacing the target label); replace it with `object_instance` in queries.
`windows_thermalzone_percent_passive_limit` and `windows_thermalzone_throttle_reasons` keep their names.
The temperature is exported in Kelvin, because `performancecounter` doesn't convert units. Replace `windows_thermalzone_temperature_celsius` in queries with:
@@ -353,6 +354,14 @@ Retired: `windows_net_route_info` was documented but never exported. It was remo
windows_exporter 0.32 follows the Prometheus naming conventions for every metric: values are in base units (seconds, bytes, hertz, ratios from 0 to 1), the unit is part of the name, and counters end in `_total`.
The metrics below were renamed. **The old names are no longer exported, there are no aliases.** Update queries, recording rules, alerts and dashboards; the sample dashboards in [`dashboard/`](dashboard/README.md) and the rules in [`contrib/mixin`](contrib/mixin) already use the new names.
### Compatibility recording rules
To migrate step by step, load [`docs/example-0.32-compatibility-rules-new-to-old.yml`](docs/example-0.32-compatibility-rules-new-to-old.yml) into Prometheus.
Its recording rules publish the metrics of 0.32 under their 0.31 names, with the 0.31 units and [instance labels](#instance-label-names), so existing dashboards and alerts keep working while you update them.
A 0.31 `instance` label comes back as `exported_instance`, the name under which Prometheus stored it.
The rules don't cover metrics that only renamed their instance label, such as `windows_service_*`, because a rule with the same name would duplicate every series, nor metrics whose 0.31 value can't be computed from 0.32.
Remove the rules once your queries use the new names.
### Unit changes
These metrics changed their value, not only their name. A graph that mixes 0.31 and 0.32 data shows a step at the upgrade.
@@ -684,6 +693,49 @@ If Prometheus scrapes with OpenMetrics, it already stored exactly these names, b
| `windows_nps_accounting_responses` | `windows_nps_accounting_responses_total` | Same value |
| `windows_nps_accounting_unknown_type` | `windows_nps_accounting_unknown_type_total` | Same value |
## Instance label names
The label that identifies the instance of a metric is now named after the entity it identifies, for example `service`, `queue` or `node`.
0.31 used the generic keys `name` and `instance` for many of them. `instance` also collided with the target label that Prometheus adds, so Prometheus renamed it to `exported_instance` (or overwrote the target label with `honor_labels: true`).
Label values are unchanged. **The old keys are no longer exported, there are no aliases.** Update label matchers, `by (...)` and `on (...)` clauses, legends and alert templates such as `{{ $labels.name }}`.
| Collector | Metrics | 0.31 | 0.32 |
|------------------|------------------------------------------------------------|-----------------------|------------------|
| `dfsr` | `windows_dfsr_connection_*` | `name` | `connection` |
| `dfsr` | `windows_dfsr_folder_*` | `name` | `folder` |
| `dfsr` | `windows_dfsr_volume_*` | `name` | `volume` |
| `dhcp` | `windows_dhcp_scope_info` | `name` | `scope_name` |
| `diskdrive` | `windows_diskdrive_*` | `name` | `device_id` (the `name` label of `windows_diskdrive_info` is removed; it repeated `device_id`) |
| `dns` | `windows_dns_wmi_stats_total` | `name` | `statistic` |
| `exchange` | `windows_exchange_ldap_*` | `name` | `process` |
| `exchange` | `windows_exchange_http_proxy_*` | `name` | `protocol` |
| `exchange` | `windows_exchange_transport_queues_*` | `name` | `queue` |
| `exchange` | `windows_exchange_workload_*` | `name` | `workload` |
| `hyperv` | `windows_hyperv_virtual_smb_*` | `instance` | `vm` |
| `mscluster` | `windows_mscluster_cluster_*` | `name` | `cluster` |
| `mscluster` | `windows_mscluster_network_*` | `name` | `network` |
| `mscluster` | `windows_mscluster_node_*` | `name` | `node` |
| `mscluster` | `windows_mscluster_resource_*` | `name`, `node_name`, `owner_group` | `resource`, `node`, `resource_group` |
| `mscluster` | `windows_mscluster_resourcegroup_*` | `name`, `node_name` | `resource_group`, `node` |
| `mscluster` | `windows_mscluster_shared_volumes_*` | `name` | `shared_volume` |
| `msmq` | `windows_msmq_*` | `name` | `queue` |
| `registry` | `windows_registry_key_success` | `name` | `key` |
| `service` | `windows_service_*` | `name` | `service` |
| `storage_spaces` | `windows_storage_spaces_pool_*` | `name` | `pool` |
| `storage_spaces` | `windows_storage_spaces_virtual_disk_*` | `name` | `virtual_disk` |
| `wmi` | `windows_wmi_query_success`, `windows_wmi_query_duration_seconds` | `name` | `query` |
| `performancecounter` | objects without `instance_label` | `instance` (stored by Prometheus as `exported_instance`, or as `instance` with `honor_labels: true`) | `object_instance` |
The `node_name` label of `windows_mscluster_resource_owner_node` and `windows_mscluster_resourcegroup_owner_node` is now `node`, like the label of the node metrics, so the owner can be joined with `on (node)`. The `owner_group` label of the resource metrics is now `resource_group`, like the label of the resource group metrics.
Info attributes that are not the instance key keep their names, for example the `name` label of `windows_cpu_info` (the processor model) and of `windows_gpu_info` (the adapter name).
For example, the stopped automatic services query becomes:
```promql
windows_service_state{state="running"} == 0
and on (instance, service) windows_service_start_mode{start_mode="auto"} == 1
```
## Corrected values
These metrics keep their names, but 0.31 reported wrong values. Graphs show a step at the upgrade.
@@ -733,7 +785,7 @@ Storage pool and virtual disk metrics are in the new `storage_spaces` collector
It works on standalone hosts too and doesn't need a failover cluster.
The `mscluster` collector now reads cluster data through the Windows Cluster API (`clusapi.dll`) instead of the `root\MSCluster` WMI namespace.
Labels and configuration are unchanged; metrics in minutes, milliseconds, hours, kilobytes, megabytes or percent are now in base units, see [Base units and counter names](#base-units-and-counter-names). `windows_mscluster_resource_is_alive_poll_interval_seconds` and `windows_mscluster_resource_looks_alive_poll_interval_seconds` are left out while the resource uses the interval of its resource type (0xFFFFFFFF in Windows). A property that can't be read is left out and reported as a collection error, instead of being exported as `0`.
Configuration is unchanged. The `name`, `node_name` and `owner_group` labels were renamed, see [Instance label names](#instance-label-names). Metrics in minutes, milliseconds, hours, kilobytes, megabytes or percent are now in base units, see [Base units and counter names](#base-units-and-counter-names). `windows_mscluster_resource_is_alive_poll_interval_seconds` and `windows_mscluster_resource_looks_alive_poll_interval_seconds` are left out while the resource uses the interval of its resource type (0xFFFFFFFF in Windows). A property that can't be read is left out and reported as a collection error, instead of being exported as `0`.
## Scrape timeouts