CDV ❯ Clustered ehcache write behind isn't properly scoped
-
Bug
-
Status: Closed
-
2 Major
-
Resolution: Fixed
-
-
-
kkannaiy
-
Reporter: gbevin
-
March 04, 2010
-
0
-
Watchers: 1
-
February 12, 2013
-
March 04, 2010
Description
The clustered write behind is scoped at cache manager level, which is a bug. I forgot to append the cache’s name to the root name for registration of the clustered async coordinator.
This affects people using different caches with write behind in the same cache manager, the clustered write behind will use the config settings of the first cache that is registered and the elements of the caches will all be processed by the same write behind queue.
Comments
Tim Eck 2010-03-04
Geert Bevin 2010-03-04
We definitely need to rework TerracottaClusteredInstanceFactory for the SPI/API stuff
Geert Bevin 2010-03-04
There’s no real workaround, people that want to isolate cache writers amongst caches will have to put the caches in different cache managers.
Fiona OShea 2010-03-04
Assigning to darwin to document as known issue. Will then re-target the jira
Fiona OShea 2010-03-05
Added Release Note http://www.terracotta.org/confluence/display/docs/Release+Notes
Kalai Kannaiyan 2010-05-10
Verified with rev 20985 tim-ehcache-2.0-system-tests/trunk and rev 21056/tc-3.2.
Kalai Kannaiyan 2010-05-10
and verified it is added to the 3.2.1 release notes
I’ll have to look at that code but maybe we should only have one map root indexed by cache manager name and keep all the sub-objects (store-map, serializers, write behind, etc) all be in the map by cache manager name. Maybe this would avoided the scope problem in the first place. Might make the code less repetitive too :-)