You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@qpid.apache.org by "Keith Wall (JIRA)" <ji...@apache.org> on 2016/10/27 14:10:58 UTC
[jira] [Comment Edited] (QPID-7472) [Java Broker, BDBStore] Removal
of large number of messages on broker startup does not free the disk space
occupied by empty BDB log files left after message removal when
overfull/undefull thresholds are not set
[ https://issues.apache.org/jira/browse/QPID-7472?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15611999#comment-15611999 ]
Keith Wall edited comment on QPID-7472 at 10/27/16 2:10 PM:
------------------------------------------------------------
Currently, the operator can force the log to be cleaned by invoke the BDBEnvironmentContainer#cleanLog managed operation.
We said:
* After synchronous recovery, which may purge transient messages that have flowed to disk, we should force a BDB cleanLog to ensure that disk utilisation shrinks in a timely way.
was (Author: k-wall):
We said:
* After synchronous recovery, which may purge transient messages that have flowed to disk, we should force a BDB cleanLog to ensure that disk utilisation shrinks in a timely way.
> [Java Broker, BDBStore] Removal of large number of messages on broker startup does not free the disk space occupied by empty BDB log files left after message removal when overfull/undefull thresholds are not set
> -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
>
> Key: QPID-7472
> URL: https://issues.apache.org/jira/browse/QPID-7472
> Project: Qpid
> Issue Type: Bug
> Components: Java Broker
> Reporter: Alex Rudyy
> Fix For: Future
>
>
> BDB Message store needs a mechanism to force cleanup of empty log files after removal of large number of messages. When flow is blocked due to exceeding of disk space utilization(by default 90%), the message removal might not help to free the disk space when underfull size and store overfull size are not set.
> In my test scenario I published transient messages which were flown to disk until disk space was consumed and flow got blocked. Then, I restarted the Broker. Flown to disk transient messages were deleted on Broker restart, but store did not shrink in size as no bdb log file was deleted. I could not publish any new message as producer flow was blocked. There was nothing to consume either as there was no persistent messages published before. As result, broker ended-up in unusable state. The following broker shutdown cleaned the disk space due to store cleanup being forced on message store close.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@qpid.apache.org
For additional commands, e-mail: dev-help@qpid.apache.org