[YUNIKORN-970] Add queue metrics with queue names as labels (#365)

Reverted as part of the YuniKorn 1.4 release steps and re-published after versioning documentation under a new commit to master. Co-authored-by: William Tom <[email protected]>
apache · Nov 20, 2023 · 2a63686 · 2a63686
1 parent 778ca9e
commit 2a63686
Showing 1 changed file with 54 additions and 7 deletions.
diff --git a/docs/metrics/queue.md b/docs/metrics/queue.md
@@ -22,13 +22,37 @@ specific language governing permissions and limitations
 under the License.
 -->
 
-## Queue
-### Application
-Eech queue has a `<queue_name> queue_app` metric to trace the applications in the queue.
+## Application
+Each queue has a `<queue_name> queue_app` metric to trace the applications in the queue.
 `<queue_name> queue_app` metrics records the number of applications in different states.
 These application states include `running`, `accepted`, `rejected`, `failed` and `completed`. 
-`<queue_name> queue_app` metrics record container states including `released`, `allocated`. 
-**Metric Type**: `guage`
+`<queue_name> queue_app` metrics record container states including `released`, `allocated`.
+
+Prior to `1.5.0` queue metrics were pushed to a separate subsystem for each queue. In `1.5.0` 
+fixed metrics with `<queue name>` labels were introduced. Metrics using `<queue name>` subsystems 
+will be **deprecated in `1.6.0` and removed in `1.7.0`** in favor of a `<queue name>` label based approach.
+
+### Label
+(Introduced in `1.5.0`)
+
+**Metric Type**: `gauge`
+
+**Namespace**: `yunikorn`
+
+**Label**: `queue: <queue name>`
+
+**TYPE**: `yunikorn_queue_app`
+
+```json
+yunikorn_queue_app{queue="root.default",state="accepted"} 3
+yunikorn_queue_app{queue="root.default",state="running"} 3
+```
+
+### Subsystem
+|:exclamation: To be deprecated in `1.6.0` and removed in `1.7.0`|
+|----------------------------------------------------------------|
+
+**Metric Type**: `gauge`
 
 **Namespace**: `yunikorn`
 
@@ -41,11 +65,34 @@ yunikorn_root_default_queue_app{state="accepted"} 3
 yunikorn_root_default_queue_app{state="running"} 3
 ```
 
-### Resource
+## Resource
 The `<queue_name> queue_resource` metric to trace the resource in the queue.
 These resource states include `guaranteed`, `max`, `allocated`, `pending`, `preempting`.
+### Label
+(Introduced in `1.5.0`)
+
+**Metric Type**: `gauge`
+
+**Namespace**: `yunikorn`
+
+**Label**: `queue: <queue name>`
+
+**TYPE**: `yunikorn_queue_resource`
+
+```json
+yunikorn_queue_resource{queue="root",resource="ephemeral-storage",state="max"} 9.41009558e+10
+yunikorn_queue_resource{queue="root",resource="hugepages-1Gi",state="max"} 0
+yunikorn_queue_resource{queue="root",resource="hugepages-2Mi",state="max"} 0
+yunikorn_queue_resource{queue="root",resource="memory",state="max"} 1.6223076352e+10
+yunikorn_queue_resource{queue="root",resource="pods",state="max"} 110
+yunikorn_queue_resource{queue="root",resource="vcore",state="max"} 8000
+```
+
+### Subsystem
+|:exclamation: To be deprecated in `1.6.0` and removed in `1.7.0`|
+|----------------------------------------------------------------|
 
-**Metric Type**: `guage`
+**Metric Type**: `gauge`
 
 **Namespace**: `yunikorn`