[
https://issues.apache.org/jira/browse/SOLR-14425?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17383717#comment-17383717
]
Mark Robert Miller edited comment on SOLR-14425 at 7/20/21, 3:20 AM:
---------------------------------------------------------------------
It's unlikely Solr tests would ever pick up any issue around using sync() or
not given that we run tests with a single ZK instance and sync is for the case
when client1 writes something to zk, communicates to client2 it should now find
that write in ZK, and client2 does a read on a different ZK instance that does
not yet see the update on it's read. It's a workaround for zk not providing
linearizable reads (as it does writes), which is done for much better read
performance. But it's not likely to matter one way or another with a single zk
instance as your zk cluster.
was (Author: markrmiller):
It's unlikely Solr tests would ever pick up any issue around using sync() or
not given that we run tests with a single ZK instance and sync is for the case
when client1 writes something to zk, communicates to client2 it should not find
that in ZK, and client2 does a read on a different ZK instance that does not
yet see the update on it's read. It's a workaround for zk not providing
linearizable reads (as it does writes), which is done for much better read
performance. But it's not likely to matter one way or another with a single zk
instance as your zk cluster.
> Fix ZK sync usage to be synchronous (blocking)
> ----------------------------------------------
>
> Key: SOLR-14425
> URL: https://issues.apache.org/jira/browse/SOLR-14425
> Project: Solr
> Issue Type: Bug
> Components: SolrCloud
> Reporter: David Smiley
> Priority: Minor
> Time Spent: 0.5h
> Remaining Estimate: 0h
>
> As of this writing, we only use one call to ZK's "sync" method. It's related
> to collection aliases -- I added this. I discovered I misunderstood the
> semantics of the API; it syncs in the background and thus returns
> immediately. Looking at ZK's sync CLI command and Curator both made me
> realize my folly. I'm considering this only a "minor" issue because I'm not
> sure I've seen a bug from this; or maybe I did in spooky test failures over a
> year ago -- I'm not sure. And we don't use this pervasively (yet).
> It occurred to me that if Solr embraced the Curator framework abstraction
> over ZooKeeper, I would not have fallen into that trap. I'll file a separate
> issue for that.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@solr.apache.org
For additional commands, e-mail: issues-h...@solr.apache.org
