[
https://issues.apache.org/jira/browse/HAMA-307?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Edward J. Yoon updated HAMA-307:
--------------------------------
Fix Version/s: (was: 0.2.1)
(was: 0.3.0)
> BSPMaster - job ID counter is not read and updated atomically
> -------------------------------------------------------------
>
> Key: HAMA-307
> URL: https://issues.apache.org/jira/browse/HAMA-307
> Project: Hama
> Issue Type: Bug
> Components: bsp
> Affects Versions: 0.2.0
> Reporter: Filipe Manana
> Assignee: Edward J. Yoon
> Fix For: 0.2.0
>
> Attachments: hama-307.patch
>
>
> The nextJobId (int attribute) in BSPMaster is not read and updated
> atomically, therefore it's possible that concurrent requests get the same job
> ID and/or cause the job ID to not get incremented by the proper quantity.
> The following simple patch fixes it.
> Notice:
> I changed the type from int to Integer.
> This is to avoid doing the following in getNewJobId():
> int id;
> synchronized (this) {
> id = nextJobId++;
> }
> the instance (referenced by this) is synchronized in many places in
> BSPMaster, so by making nextJobId an Integer (object), we can synchronize on
> it, reducing contention times:
> private Integer nextJobId = Integer.valueOf(1);
> public BSPJobID getNewJobId() throws IOException {
> int id;
> synchronized (nextJobId) {
> id = nextJobId;
> nextJobId = Integer.valueOf(id + 1);
> }
> return new BSPJobID(this.masterIdentifier, id);
> }
> cheers
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
