[
https://issues.apache.org/jira/browse/ARTEMIS-2244?focusedWorklogId=199072&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-199072
]
ASF GitHub Bot logged work on ARTEMIS-2244:
-------------------------------------------
Author: ASF GitHub Bot
Created on: 15/Feb/19 01:42
Start Date: 15/Feb/19 01:42
Worklog Time Spent: 10m
Work Description: wy96f commented on issue #2533: ARTEMIS-2244
checkDepage method placed outside CRITICAL_DELIVER avoid critical analyzer
timeout
URL: https://github.com/apache/activemq-artemis/pull/2533#issuecomment-463874114
> To me it seems we shouldnt have change critical analyslzer at all. Its
done its job. Its detected on a critical path the is something blocking and it
shouldnt be. The real fix is to fix the block, not remove the safety check.
>
> It was added for this exact purpose!
Agreed. To fix the blocking problem, could we add a flag indicating
checkDepage() is in progress? If in process, we could return directly, no need
to block in iterator.hasNext() which potentially causes triggering critical
analyzing.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
Issue Time Tracking
-------------------
Worklog Id: (was: 199072)
Time Spent: 3h (was: 2h 50m)
> checkDepage method placed outside CRITICAL_DELIVER avoid critical analyzer
> timeout
> ----------------------------------------------------------------------------------
>
> Key: ARTEMIS-2244
> URL: https://issues.apache.org/jira/browse/ARTEMIS-2244
> Project: ActiveMQ Artemis
> Issue Type: Bug
> Reporter: yuebao
> Priority: Major
> Attachments: threaddump.txt
>
> Time Spent: 3h
> Remaining Estimate: 0h
>
> We found server crash becauseof critical analyzer timeout, thread dump can be
> seen in attachment.
> Suppose that there is a topic t with two subscriber ta and tb. Some messages
> were routed to subscriber ta, not to tb. When a consumer that subscribe tb is
> created and send ConsumerCredits to server, ServerConsumerImpl will call
> promptDelivery method, then forceDelivery() -> messageQueue.deliverAsync() ->
> deliverRunner will executed in the pageSubscription's executor(step1), then
> checkDepage(step2) -> pageIterator.hasNext(synchronized method) -> next ->
> moveNext, moveNext method will iterate through queue until find a matching
> message. At this time(step1), DeliverRunner call deliver method ->
> checkDepage -> pageIterator.hasNext, this step blocked on CursorIterator
> which was locked by step2, then critical analyzer timeout.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
