Details
-
Bug
-
Status: Closed
-
Major
-
Resolution: Duplicate
-
5.9.0
-
None
-
None
-
Windows Server 2012
Description
We have a cluster setup using "Shared File System Master Slave". The kahadb database is stored on a 20 GB shared filesystem. The last weeks we have had a couple of times where producers where stopped because the persistent store usage was at 100%. I found that where about 8 GB of disk available, so the persistent store usage shouldn't be more than around 60%.
I found that when starting up or failing over the max size of the persistent store is adjusted to be no more then the available disk space. Found this in the log:
INFO | jvm 1 | 2014/07/27 06:36:09 | WARN | Store limit is 20480 mb, whilst the data directory: \\pulbrscbfs1\ActiveMQ\kahadb only has 7945 mb of usable space - resetting to maximum available disk space: 7945 mb
And then this occurs:
INFO | jvm 1 | 2014/07/27 06:40:03 | INFO | Usage(default:store:queue://PCIS.SCH.Notes.Jobs:store) percentUsage=153%, usage=12813873606, limit=8331595776, percentUsageMinDelta=1%;Parent:Usage(default:store) percentUsage=153%, usage=12813873606, limit=8331595776, percentUsageMinDelta=1%: Persistent store is Full, 100% of 8331595776. Stopping producer (ID:PULBRSCBMULE2-49430-1406435872187-1:1:4:1) to prevent flooding queue://PCIS.SCH.Notes.Jobs. See http://activemq.apache.org/producer-flow-control.html for more info (blocking for: 2s)
As you can see the percentUsage is 153%!
So the reset of the store limit to 7945 mb is wrong and should have been 12220 mb + 7945 mb = 20165 mb
Attachments
Issue Links
- duplicates
-
AMQ-4832 Broker limits store space, not taking into account what it's already using
- Resolved