[
https://issues.apache.org/jira/browse/GIRAPH-246?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Eli Reisman updated GIRAPH-246:
-------------------------------
Attachment: GIRAPH-246-NEW-FIX-2.patch
This includes one more extra call to progress() and is otherwise just like the
first "NEW-FIX" patch.
Neither trunk nor any of these PredicateLock fixes allow std err and std out
logs reach the HTML Mapper Detail pages for some reason, the old rebase does.
The only times my old rebase blew up and "timed out" today is when Netty blew
up under stress tests, as the Mapper Details revealed.
This had "timeouts" occasionally in the same circumstances but there was no log
to confirm why (which is why I doubted the earlier "new fix" patch.) We can
commit this and hope it was just Netty blowing up during stress tests, if we
insist on using a PredicateLock solution. This as closely replicates the good
behavior of the old rebase patch as any I've tested yet.
Alternately, I can upload the latest rebase of the original 246 and we can
commit that and wait for a better fix, but if you guys are having trouble
replicating this problem then chances are shooting in the dark for a fix to
replace the rebase will not happen, and I know you don't like seeing the
progress calls in BspServiceWorker.
Either way is fine at this point, I will be running a lot of code over the next
few days/nights as I get clear windows to do so, but I am 100% sure after
running trunk a bunch today as I was doing A/B tests on these that the current
fix does not work for us here. I'd love to pick one (the rebase or this) and
commit, its at least a step forward from what we have which is just not
functioning for us at all. users here are getting sick of manually adding
patches to run Giraph on our job queue ;)
> Periodic worker calls to context.progress() will prevent timeout on some
> Hadoop clusters during barrier waits
> -------------------------------------------------------------------------------------------------------------
>
> Key: GIRAPH-246
> URL: https://issues.apache.org/jira/browse/GIRAPH-246
> Project: Giraph
> Issue Type: Improvement
> Components: bsp
> Affects Versions: 0.2.0
> Reporter: Eli Reisman
> Assignee: Eli Reisman
> Priority: Minor
> Labels: hadoop, patch
> Fix For: 0.2.0
>
> Attachments: GIRAPH-246-10.patch, GIRAPH-246-11.patch,
> GIRAPH-246-1.patch, GIRAPH-246-2.patch, GIRAPH-246-3.patch,
> GIRAPH-246-4.patch, GIRAPH-246-5.patch, GIRAPH-246-6.patch,
> GIRAPH-246-7.patch, GIRAPH-246-7_rebase1.patch, GIRAPH-246-8.patch,
> GIRAPH-246-9.patch, GIRAPH-246-NEW-FIX-2.patch, GIRAPH-246-NEW-FIX.patch
>
>
> This simple change creates a command-line configurable option in GiraphJob to
> control the time between calls to context().progress() that allows workers to
> avoid timeouts during long data load-ins in which some works complete their
> input split reads much faster than others, or finish a super step faster. I
> found this allowed jobs that were large-scale but with low memory overhead to
> complete even when they would previously time out during runs on a Hadoop
> cluster. Timeout is still possible when the worker crashes or runs out of
> memory or has other GC or RPC trouble that is legitimate, but prevents
> unintentional crashes when the worker is actually still healthy.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
