[
https://issues.apache.org/jira/browse/SOLR-12833?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16687573#comment-16687573
]
Mark Miller commented on SOLR-12833:
------------------------------------
Hey [~yuanyun.cn], I was looking through the patch and I have some thoughts on
some slight renaming, but I can handle that, otherwise it's looking okay.
One question I had was when try lock fails - don't we still go down the logic
as if we had gotten the lock?
{code:java}
vBucketLocked = tryGetVersionBucketLock(bucket);
bucket.wakeUpAll();
//just in case anyone is waiting let them know that we have a new update
// we obtain the version when synchronized and then do the add so we can
ensure that
// if version1 < version2 then version1 is actually added before version2.
// even if we don't store the version field, synchronizing on the bucket
// will enable us to know what version happened first, and thus enable
// realtime-get to work reliably.
// TODO: if versions aren't stored, do we need to set on the cmd anyway
for some reason?
// there may be other reasons in the future for a version on the commands
if (versionsStored) {
{code}
> Use timed-out lock in DistributedUpdateProcessor
> ------------------------------------------------
>
> Key: SOLR-12833
> URL: https://issues.apache.org/jira/browse/SOLR-12833
> Project: Solr
> Issue Type: Improvement
> Security Level: Public(Default Security Level. Issues are Public)
> Components: update, UpdateRequestProcessors
> Affects Versions: 7.5, master (8.0)
> Reporter: jefferyyuan
> Assignee: Mark Miller
> Priority: Minor
> Fix For: master (8.0)
>
>
> There is a synchronize block that blocks other update requests whose IDs fall
> in the same hash bucket. The update waits forever until it gets the lock at
> the synchronize block, this can be a problem in some cases.
>
> Some add/update requests (for example updates with spatial/shape analysis)
> like may take time (30+ seconds or even more), this would the request time
> out and fail.
> Client may retry the same requests multiple times or several minutes, this
> would make things worse.
> The server side receives all the update requests but all except one can do
> nothing, have to wait there. This wastes precious memory and cpu resource.
> We have seen the case 2000+ threads are blocking at the synchronize lock, and
> only a few updates are making progress. Each thread takes 3+ mb memory which
> causes OOM.
> Also if the update can't get the lock in expected time range, its better to
> fail fast.
>
> We can have one configuration in solrconfig.xml:
> updateHandler/versionLock/timeInMill, so users can specify how long they want
> to wait the version bucket lock.
> The default value can be -1, so it behaves same - wait forever until it gets
> the lock.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
