Print

Print


This is the content of my /nfs/atlas/local/lib

lrwxrwxrwx 1 sgmatl034 atlsgm 52 Jul 17 23:11 /nfs/atlas/local/lib/libglobus_gssapi_gsi_gcc32dbgpthr.so -> /opt/globus/lib/libglobus_gssapi_gsi_gcc32dbgpthr.so
lrwxrwxrwx 1 sgmatl034 atlsgm 52 Oct 28 02:30 /nfs/atlas/local/lib/libglobus_gssapi_gsi.so -> /opt/globus/lib/libglobus_gssapi_gsi_gcc32dbgpthr.so
lrwxrwxrwx 1 sgmatl018 atlsgm 22 Nov 28 02:10 /nfs/atlas/local/lib/libshift.so -> /opt/lcg/lib/libdpm.so
lrwxrwxrwx 1 sgmatl018 atlsgm 22 Nov 28 02:10 /nfs/atlas/local/lib/libshift.so.2.1 -> /opt/lcg/lib/libdpm.so

all the links are void, they point to files that don't exist anymore and the jobs don't fail in Manchester. AFAIK the jobs only source the local setup file which for most sites is standard. Mine looks like this

cat /nfs/atlas/local/setup.sh
# Local setup
export ATLAS_POOLCOND_PATH="/cvmfs/atlas.cern.ch/repo/conditions"
export FRONTIER_SERVER="(serverurl=http://lcgft-atlas.gridpp.rl.ac.uk:3128/frontierATLAS)(serverurl=http://ccfrontier.in2p3.fr:23128/ccin2p3-AtlasFrontier)(proxyurl=http://squid-cache.tier2.hep.manchester.ac.uk:3128)(proxyurl=http://fal-pygrid-45.lancs.ac.uk:3128)"
export FRONTIER_LOG_LEVEL=warning
#export FRONTIER_READTIMEOUTSECS=60
if [ -f /etc/emi-version ] ; then
    export LD_LIBRARY_PATH=/nfs/atlas/local/emi/lib:/nfs/atlas/local/emi/lib64:$LD_LIBRARY_PATH
else
    export LD_LIBRARY_PATH=/nfs/atlas/local/lib:/nfs/atlas/local/lib64:$LD_LIBRARY_PATH
fi
# allow local override at end
[ -f /nfs/atlas/local/setup.sh.local ] && source /nfs/atlas/local/setup.sh.local


and infact if I look at /nfs/atlas/local/emi I find the link to the correct libraries

ls -l /nfs/atlas/local/emi/*
/nfs/atlas/local/emi/lib:
total 0
lrwxrwxrwx 1 sgmatl018 atlsgm 24 Nov 28 02:10 libshift.so -> /usr/lib/libdpm.so.1.8.4
lrwxrwxrwx 1 sgmatl018 atlsgm 24 Nov 28 02:10 libshift.so.2.1 -> /usr/lib/libdpm.so.1.8.4

/nfs/atlas/local/emi/lib64:
total 0
lrwxrwxrwx 1 sgmatl018 atlsgm 20 Nov 28 02:10 libshift.so -> /usr/lib64/libdpm.so
lrwxrwxrwx 1 sgmatl018 atlsgm 20 Nov 28 02:10 libshift.so.2.1 -> /usr/lib64/libdpm.so

so I think Alessandro's jobs haven't landed yet on one of your EMI WNs. So either you setup a separate queue or you reinstall everything as EMI.

cheers
alessandra


On 29/11/2012 09:48, Mark Slater wrote:
[log in to unmask]" type="cite">
Hi Alesandra,

But how would the atlas file get changed? Does it get picked up automatically or do I have to tell Alessandro De Salvo to send an install job? As it stands, if I shifted everything over then the jobs would still fail because the links in this atlas dir point to nothing!

Maybe there's some additional setup I'm missing? In EMI2 Globus doesn't seem to be in /opt anymore....

THanks,

Mark

On 29/11/12 09:45, Alessandra Forti wrote:
[log in to unmask]" type="cite">
Hi Mark,

RAL has run WNs in parallel however different setups might fail. In your case I'd really just reinstall the whole lot in one go.

cheers
alessandra

On 29/11/2012 09:28, Mark Slater wrote:
[log in to unmask]" type="cite">
Hi All,

Just forwarding my atlas cloud support question to see if anyone else had the same problem! I would prefer to start shifting all workers over to EMI2 ASAP!

Thanks,

Mark

-------- Original Message --------
Subject: Question regarding EMI2 WN
Date: Wed, 28 Nov 2012 12:20:20 +0000
From: Mark Slater <[log in to unmask]>
To: atlas-support-cloud-uk (ATLAS support contact for UK cloud) <[log in to unmask]>


Hiya,

I've updated a couple of Bham's nodes to EMI2 but it seems we're failing 
prod jobs on them because of some bad links to libraries. If I do:

[root@epgf01 ~]#   ls /egee/soft/atlas/cvmfs/local/lib/ -ltr
total 12
lrwxrwxrwx 1 sgmatl12 atlassgm 22 May 21  2012 libshift.so -> 
/opt/lcg/lib/libdpm.so
lrwxrwxrwx 1 sgmatl12 atlassgm 22 May 21  2012 libshift.so.2.1 -> 
/opt/lcg/lib/libdpm.so
lrwxrwxrwx 1 sgmatl12 atlassgm 52 May 21  2012 
libglobus_gssapi_gsi_gcc32dbgpthr.so -> 
/opt/globus/lib/libglobus_gssapi_gsi_gcc32dbgpthr.so


the globus libs are now no longer here and in fact, I believe might be 
in /usr/share...

Is there anything extra I need to do to fix this or can I only run Glite 
3.2 WN OR EMI2 WN, not both at the same time?

Many Thanks,

Mark

P.S. A typical failed job:

http://panda.cern.ch/server/pandamon/query?job=1662074376




-- 
Facts aren't facts if they come from the wrong people. (Paul Krugman)



-- 
Facts aren't facts if they come from the wrong people. (Paul Krugman)