[jira] [Commented] (AMQ-5196) Corrupted Kahadb store after failover: "Could not load journal record. Invalid location"

["Matt Pavlovich (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/AMQ-5196?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17244617#comment-17244617 ] 

Matt Pavlovich commented on AMQ-5196:
-------------------------------------

[~windeyu] What was the shared filesystem used in this? I suspect if it was Windows-based SMB/CIFS that it was straight-up corruption from Windows handing both instances a lock. 

Is this still an issue with more recent versions of ActiveMQ? 5.15.14 or 5.16.0

Notes: Using CIFS or SMB file shares on Windows servers is not recommended, since the locking mechanism is not robust on that OS.

Please note, this JIRA is marked for close-pending if there are no updated in 30-days.

> Corrupted Kahadb store after failover: "Could not load journal record. Invalid location"
> ----------------------------------------------------------------------------------------
>
>                 Key: AMQ-5196
>                 URL: https://issues.apache.org/jira/browse/AMQ-5196
>             Project: ActiveMQ
>          Issue Type: Bug
>          Components: Broker, Message Store
>    Affects Versions: 5.9.0
>         Environment: Windows 7.
> <persistenceAdapter>
>     <kahaDB directory="\\some.network.lan\amq" ignoreMissingJournalfiles="true" checksumJournalFiles="true" checkForCorruptJournalFiles="true"/>
> </persistenceAdapter>
>            Reporter: Edwin Yu
>            Priority: Major
>         Attachments: masterNode.log, slaveNode.log
>
>
> I'm not sure if I can reproduce it, but it's important for me to report.  We have a master/slave setup sharing the kahaDB store on a distributed network drive.  The running master encountered some intermittent  network issue for ten minutes or so, and finally invoked my DefaultIOExceptionHandler, which shut down the JVM.  The slave took over and started up.  All remote consumers reconnected to the slave node via the failover protocol.  However, the slave node repeatedly complained about "Could not load journal record. Invalid location".  All consumers were idle and didn't process any messages.  Logs from both nodes are attached.
> We're running ActiveMQ 5.9.0.  I copied over the kabadb folder to the latest ActiveMQ 5.9.1 where I also set ignoreMissingJournalfiles="true" checksumJournalFiles="true" checkForCorruptJournalFiles="true".  It started up with the same error.  It indicated to me that 5.9.1 cannot recover from the corrupted store either.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)