On 16/05/14 16:59, Kashif Mohammad wrote:
> Hi Chris
>
> I have analysed pepd access log of one hour and found that on average it is receiving 60-200 requests per minute and only around 5% requests are coming from glexec. I am surprised by the high number of requests from CE's considering that we have only 1500 job slots.
> Did you see this kind of behaviour?
> I will try to look into CE's log to as well.
in /var/log/argus/pepd:
zgrep ACCOUNTMAPPER_OH process-2014-05-13.0.log.gz | wc -l
200985
Which is about 140 requests a minute on average!!!
All our CEs use argus for authentication.
I haven't worked out the glexec/CE split.
Chris
> Thanks
> Kashif
>
>
>> -----Original Message-----
>> From: Testbed Support for GridPP member institutes [mailto:TB-
>> [log in to unmask]] On Behalf Of Chris Brew
>> Sent: 16 May 2014 15:28
>> To: [log in to unmask]
>> Subject: Re: ARGUS problem
>>
>> Hi,
>>
>> Installing a second Argus and pointing our ArcCEs at that fixed the problem
>> for a while. It was running fine handling those until I wanted to rebuild the
>> first which was still having problems doing the mappings for gLexec. When I
>> switched that over to the new server if fell over in a heap.
>>
>> So I've switched the ArcCEs back over to the original (but reinstalled) Argus
>> and they are working fine but the second box is now mostly failing to map for
>> gLexec - if a few service restarts and maybe reboots doesn't fix that I'll try to
>> reinstall (VMs and puppet for teh win) and see if a fresh Argus server can
>> handle the gLexec load.
>>
>> Yours,
>> Chris.
>>
>> On 16/05/2014 10:01, "Kashif Mohammad"
>> <[log in to unmask]<mailto:[log in to unmask]
>> C.UK>> wrote:
>>
>> Hi
>>
>> Our Argus server is behaving strangely in last few days. We never had any
>> issue with Argus server since beginning and the last time we updated emi
>> packages were more than a month ago. Every few hour it just refuses to
>> entertain any request for mapping and requests times out. In between it
>> works perfectly ok. I have already applied most of the workaround
>> mentioned in
>>
>> https://twiki.cern.ch/twiki/bin/view/EGEE/ArgusEMIDeployment#Performa
>> nce_issue_with_Argus_PEP
>>
>> I have applied fallowing workarounds
>> Disabled caching mechanism in /etc/argus/pepd/pepd.ini Increased default
>> memory for java to 1GB in /etc/sysconfig/argus-pepd
>>
>> restarted services and rebooted as well.
>>
>> We are running argus server on a VM with 4GB RAM and 2 cpu. It is serving
>> around 1500 job slots. Ganglia plots are not showing any thing intresting.
>>
>> I remember that few other sites had same kind of issue, I will be interested
>> to know that how you solved this problem.
>>
>> Thanks
>> Kashif
>>
>>
>> --
>> Scanned by iCritical.
|