[
https://issues.apache.org/jira/browse/SOLR-12088?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16399138#comment-16399138
]
Erick Erickson commented on SOLR-12088:
---------------------------------------
Another question:
Does the latency persist longer than any system timeouts? Put another way:
If you start all the Solr instances fresh and some nodes are down, is there
still latency?
What I'm thinking of here is that it may take up until some timeout for each
Solr instance to "see" that the node is down.
For instance, if I kill a Solr node with a -9, it has no chance to tell ZK (and
ZK in turn inform the rest of the collection) that it's going away. The rest of
the collection finds out about this by one of several methods, all involving
some timeout (ZK occasionally pings, leader sends request to update etc.).
So if this is transient it may be functioning as expected, but if it lasts well
past all the possible timeouts it's another story.
> Shards with dead replicas cause increased write latency
> -------------------------------------------------------
>
> Key: SOLR-12088
> URL: https://issues.apache.org/jira/browse/SOLR-12088
> Project: Solr
> Issue Type: Bug
> Security Level: Public(Default Security Level. Issues are Public)
> Components: SolrCloud
> Affects Versions: 7.2
> Reporter: Jerry Bao
> Priority: Major
>
> If a collection's shard contains dead replicas, write latency to the
> collection is increased. For example, if a collection has 10 shards with a
> replication factor of 3, and one of those shards contains 3 replicas and 3
> downed replicas, write latency is increased in comparison to a shard that
> contains only 3 replicas.
> My feeling here is that downed replicas should be completely ignored and not
> cause issues to other alive replicas in terms of write latency.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
