On Tue, 22 Mar 2011 14:44:24 +0100
Francisco Bernabé (Paco) wrote:
> Hi,
Hi Paco,
[...]
> So far seems to be working ok, but I've seems lines like the
> following ones in the workload manager log files:
>
> 16 Mar, 21:01:23 -I: [Info] postpone(submit_request.cpp:212):
> postponing https://graskant.nikhef.nl:9000/UAlyKM8AOwjVcKGiEC_m7Q
> (BrokerHelper: no compatible resources) 16 Mar, 21:11:35 -I: [Info]
> postpone(submit_request.cpp:212): postponing
> https://grasveld.nikhef.nl:9000/IaRl4teiIzh-mShxRjhmwA (BrokerHelper:
> no compatible resources)
we had similar problem and did some modification
in /opt/glite/etc/glite_wms.conf:
1.-) IsmIILDAPSearchAsync = true
2.-) II_Timeout=300
3.-) Configure google-perftools:
RuntimeMalloc = "/usr/lib/libtcmalloc_minimal.so";
http://glite.web.cern.ch/glite/packages/R3.1/deployment/glite-WMS/glite-WMS-known-issues.asp
since we did those changes, we have not seen anymore "no compatible
resources" error.
[...]
> So, my guess is that it's skipping clusters and CEs until the proper
> ones are found, and if they're not found then the message
> "(BrokerHelper: no compatible resources)" is showed in the log file.
> Anyway, when I get this issue, the ~/new directory starts to get more
> and more full of files, and once it reaches 1500, no more jobs can be
> sent to this WMS. The workaround for this is to restart the workload
> manager, and I have created a nagios plugin to warn about the
> increase of files in ~/new, and I have changed the LogLevel to 6, in
> order to get some more information the next time it happens; but in
> the meanwhile, did anybody have this same issue? The workaround is
> pretty easy to execute, but it would be better if there was no
> problem at all.
IIRC you could modify that 1500 limit in the same cofn file:
--jdnum
but please, google it first..
> Cheers,
> Paco.
HTH,
Arnau
|