[
https://issues.jboss.org/browse/ISPN-2835?page=com.atlassian.jira.plugin....
]
Dan Berindei resolved ISPN-2835.
--------------------------------
Resolution: Cannot Reproduce Bug
This is not a problem in the test setup, not in M/R.
The test uses {{AbstractCacheTest.getDefaultClusteredCacheConfig()}} to construct the
configuration, which disables state transfer. If the cache isn't started on both nodes
before the keys are inserted in the cache, one of the nodes will not receive any entries
via state transfer and naturally the M/R task won't find any entries there.
I fixed Galder's test by ading this line in the createCacheManagers method:
{code}
builder.clustering().stateTransfer().fetchInMemoryState(true);
{code}
Issues w/ M/R test cases if cache are not explicitly started on all
nodes
-------------------------------------------------------------------------
Key: ISPN-2835
URL:
https://issues.jboss.org/browse/ISPN-2835
Project: Infinispan
Issue Type: Bug
Components: Core API, Distributed Execution and Map/Reduce
Reporter: Ray Tsang
Assignee: Dan Berindei
Labels: onboard
Fix For: 5.3.0.Final
Attachments: infinispan.log, mr-test-src.zip
I ran into some issues while using M/R. The gist of the issue I was seeing is that:
Start a cluster of Embedded Caches, like 4 nodes
Put in 100 elements
Run a simple M/R job to count the number of keys
If I run the M/R job using the node I'm inserting elements into as coordinator - the
result is 100
But if I run the M/R job using a different node as coordinator, the result is less than
100
More interestingly, I can pause for 5 seconds and run the M/R jobs again, the results are
always less than 100
This behavior doens't occur if I explicitly run cacheManager.getCache() for each of
the nodes...
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see:
http://www.atlassian.com/software/jira