Add active worker count and active coordinator count metrics #27408
Conversation
gertjanal
force-pushed
the
add_activeWorkerCount_and_activeCoordinatorCount_metrics
branch
from
c3f98f0 to
8243d5f
Compare
gertjanal
force-pushed
the
add_activeWorkerCount_and_activeCoordinatorCount_metrics
branch
2 times, most recently
from
9990048 to
55146ba
Compare
| private final boolean worker; | ||
| private final long longHashCode; | ||
|
|
||
| public InternalNode(String nodeIdentifier, URI internalUri, NodeVersion nodeVersion, boolean coordinator) |
There was a problem hiding this comment.
This constructor is used in quite some test files. Shall I remove this constructor and update the tests, or shall I keep it this way?
gertjanal
force-pushed
the
add_activeWorkerCount_and_activeCoordinatorCount_metrics
branch
from
55146ba to
110890f
Compare
| <item> | ||
| <ignore>true</ignore> | ||
| <code>java.method.addedToInterface</code> | ||
| <new>method boolean io.trino.spi.Node::isWorker()</new> |
There was a problem hiding this comment.
It's unnecessary to add it, since the worker means not coordinator. we already has isCoordinator, why not just reuse it
There was a problem hiding this comment.
Because a node can be a worker and coordinator at the same time using node-scheduler.include-coordinator as described here: https://trino.io/docs/current/installation/deployment.html#config-properties
There was a problem hiding this comment.
I think it is still a coordinator, that is allow schedule task on it but it is not a worker
There was a problem hiding this comment.
With node-scheduler.include-coordinator=true, the UI adds it to the active worker count, but that number is not available via the metrics. The only metric available is the number of all nodes in the cluster.
So with node-scheduler.include-coordinator=true, the active node count is the number of available workers, with node-scheduler.include-coordinator=false` (the default), the active node count - the number of coordinators is the number of available workers. This PR introduces a new metric that is always consistent with the number of active workers as shown in the UI
gertjanal
force-pushed
the
add_activeWorkerCount_and_activeCoordinatorCount_metrics
branch
from
110890f to
ae23a22
Compare
2 participants
Description
The
/metricsendpoint lists active nodes, but not all nodes are workers and # of workers is not always equal to # nodes - # coordinators as coordinators can be workers as well.This PR adds
CoordinatorNodeManager_ActiveCoordinatorCountandCoordinatorNodeManager_ActiveWorkerCountmetrics. I can't find any tests for the current metrics in theCoordinatorNodeManager, so I'm not sure they should be added.Release notes
( ) This is not user-visible or is docs only, and no release notes are required.
( X ) Release notes are required. Please propose a release note for me.
( ) Release notes are required, with the following suggested text: