[
https://issues.apache.org/jira/browse/SOLR-5944?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14948865#comment-14948865
]
Ishan Chattopadhyaya edited comment on SOLR-5944 at 10/8/15 3:42 PM:
---------------------------------------------------------------------
bq. Looks like this needs to be nailed in more detail. What exactly happens to
other writes when an out of order update reaches a replica?.
I think those other writes (i.e. writes for other documents) would continue to
happen, since they are being performed using separate threads. As for other
writes to the same document, I think the consequence would be that they'll keep
getting blocked until the dependent update is written to the replica (after
which they'll get unblocked one by one, unless they timeout).
bq. Milliseconds can be pushed to many seconds due to GC, etc. Also, it gets
very tricky if a replica goes down...
Do you guys think it would be reasonable to let user configure this, and
document somewhere on the lines of: if you have lots of back of back updates
happening on the same documents, then there will be a possibility of threads
waiting for other threads, and hence this timeout can be increased, say, to
seconds.
bq. Will the proposed change have any impact on the performance for users who
do not use this feature?
I think there will be no impact, since this wait would only affect in-place
updates. Unless if we want to make sure there will be no reordering in the
first place.
was (Author: ichattopadhyaya):
bq. Looks like this needs to be nailed in more detail. What exactly happens to
other writes when an out of order update reaches a replica?.
I think those other writes (i.e. writes for other documents) would continue to
happen, since they are being performed using separate threads. As for other
writes to the same document, I think the consequence would be that they'll keep
getting blocked until the dependent update is written to the replica (after
which they'll get unblocked one by one, unless they timeout).
bq. Milliseconds can be pushed to many seconds due to GC, etc. Also, it gets
very tricky if a replica goes down...
Do you guys think it would be reasonable to let user configure this, and
document somewhere on the lines of: if you have lots of back of back updates
happening on the same documents, then there will be a possibility of threads
waiting for other threads, and hence this timeout can be increased, say, to
seconds.
bq. Will the proposed change have any impact on the performance for users who
do not use this feature?
I think there will be no impact, since this wait would only affect in-place
updates.
> Support updates of numeric DocValues
> ------------------------------------
>
> Key: SOLR-5944
> URL: https://issues.apache.org/jira/browse/SOLR-5944
> Project: Solr
> Issue Type: New Feature
> Reporter: Ishan Chattopadhyaya
> Assignee: Shalin Shekhar Mangar
> Attachments: SOLR-5944.patch, SOLR-5944.patch, SOLR-5944.patch,
> SOLR-5944.patch, SOLR-5944.patch, SOLR-5944.patch, SOLR-5944.patch,
> SOLR-5944.patch, SOLR-5944.patch, SOLR-5944.patch, SOLR-5944.patch,
> SOLR-5944.patch
>
>
> LUCENE-5189 introduced support for updates to numeric docvalues. It would be
> really nice to have Solr support this.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
