[
https://issues.apache.org/jira/browse/SOLR-9858?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15890512#comment-15890512
]
Andrzej Bialecki commented on SOLR-9858:
-----------------------------------------
Comments to your comments:
1. Fixing.
2. We can make this handler smarter. I thought it would be less complicated to
allow any updates to be processed, especially since we're not sure yet what we
want to aggregate where, and also to allow for future extensibility of this
mechanism. We can lock it down later. The issue of sending reports to the node
that recently changed its role (eg. stopped being Overseer) - I don't think
it's a big deal, in the worst case we lose just one report, because the next
report will go to the right node.
3. (and 8) I've modified {{SolrClientCache}} to use the provided instance of
{{HttpClient}} and both {{SolrOverseerReporter}} and {{SolrReplicaReporter}}
use the one obtained from {{UpdateShardHandler}}.
4. After refactoring the way to construct these names stopped being so simple -
I'm moving these tests now to {{SolrCloudReportersTest}}.
5. Good point - fixing.
6. Fixing.
7. I'll make a separate issue for this.
9. This is on purpose. Imagine a collection with replication 1 - the only
replica is also a shard leader. If we prevented the leader from reporting its
state we wouldn't get any data in the "leader" registry.
10. Indeed, there seems to be a bug somewhere - investigating...
Updated patch will follow.
> Collect aggregated metrics from nodes and shard leaders in overseer
> -------------------------------------------------------------------
>
> Key: SOLR-9858
> URL: https://issues.apache.org/jira/browse/SOLR-9858
> Project: Solr
> Issue Type: Improvement
> Security Level: Public(Default Security Level. Issues are Public)
> Components: metrics
> Affects Versions: master (7.0)
> Reporter: Andrzej Bialecki
> Assignee: Andrzej Bialecki
> Priority: Minor
> Attachments: SOLR-9858.patch
>
>
> Overseer can collect metrics from Solr nodes and shard leaders in order to
> have a notion of the indexing / query / replication / system load on each
> node, shard and its replicas. This information then can be used for cluster
> (auto)scaling.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
