[
https://issues.apache.org/jira/browse/CASSANDRA-12685?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15515710#comment-15515710
]
Stefan Podkowinski commented on CASSANDRA-12685:
------------------------------------------------
The
[HintsDispatchExecutor|https://github.com/apache/cassandra/blob/cee22ad54d7c28aaf0868dd45aeb9b5b708b0e78/src/java/org/apache/cassandra/hints/HintsDispatchExecutor.java#L272]
will only remove hint files in case all entries have successfully dispatched,
else the last offset is saved that is known to have been delivered
successfully. The executer itself should be periodically run by the
[HintsDispatchTrigger|https://github.com/apache/cassandra/blob/087264f29c309ed0b2a18cc20b79016a05847f1c/src/java/org/apache/cassandra/hints/HintsDispatchTrigger.java].
> Add retry to hints dispatcher
> -----------------------------
>
> Key: CASSANDRA-12685
> URL: https://issues.apache.org/jira/browse/CASSANDRA-12685
> Project: Cassandra
> Issue Type: Improvement
> Reporter: Dikang Gu
> Assignee: Dikang Gu
> Priority: Minor
> Fix For: 3.x
>
>
> Problem: I often see timeout in hints replay, I find there is no retry for
> hints replay, I think it would be great to add some retry logic for timeout
> exception.
> {code}
> 2016-09-20_07:32:01.16610 INFO 07:32:01 [HintedHandoff:3]: Started hinted
> handoff for host: 859af100-5d45-42bd-92f5-2bc78822158b with IP:
> /2401:db00:12:30d7:face:0:39:0
> 2016-09-20_07:58:49.29983 INFO 07:58:49 [HintedHandoff:3]: Timed out
> replaying hints to /2401:db00:12:30d7:face:0:39:0; aborting (55040 delivered)
> 2016-09-20_07:58:49.29984 INFO 07:58:49 [HintedHandoff:3]: Enqueuing flush
> of hints: 15962349 (0%) on-heap, 2049808 (0%) off-heap
> 2016-09-20_08:02:17.55072 INFO 08:02:17 [HintedHandoff:1]: Started hinted
> handoff for host: 859af100-5d45-42bd-92f5-2bc78822158b with IP:
> /2401:db00:12:30d7:face:0:39:0
> 2016-09-20_08:05:45.25723 INFO 08:05:45 [HintedHandoff:1]: Timed out
> replaying hints to /2401:db00:12:30d7:face:0:39:0; aborting (7936 delivered)
> 2016-09-20_08:05:45.25725 INFO 08:05:45 [HintedHandoff:1]: Enqueuing flush
> of hints: 2301605 (0%) on-heap, 259744 (0%) off-heap
> 2016-09-20_08:12:19.92910 INFO 08:12:19 [HintedHandoff:2]: Started hinted
> handoff for host: 859af100-5d45-42bd-92f5-2bc78822158b with IP:
> /2401:db00:12:30d7:face:0:39:0
> 2016-09-20_08:51:44.72191 INFO 08:51:44 [HintedHandoff:2]: Timed out
> replaying hints to /2401:db00:12:30d7:face:0:39:0; aborting (83456 delivered)
> {code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
