FAQ CMSSW on the Grid
FAQ
- Using CRAB, I get a "VOMS extension not found" error message
It is typically what happens if you did a :
voms-proxy-init
instead of a :
voms-proxy-init --voms cms:/cms/becms
- How can I check the validity of my Grid proxy ?
voms-proxy-info --all
- How can I destroy my proxy ?
voms-proxy-destroy
- Why is my interactive analysis on a local Grid file so slow ?
- are you writing CMSSW EDProducts to file, in addition to root histograms ? If yes, this is very slow. Try to comment out the endpath in your config file and re-run.
- myproxy-info times out, forcing re-authentication each time a crab job is submitted. How can I by-pass that ?
- At IIHE, modify the [GRID] section of your crab.cfg as follows:
[GRID] dont_check_proxy = 1
- Using CRAB, I got a "ASAP ERROR: Unable to ship a valid proxy to the server"
When you do a "crab -submit", it tries to delegate proxy generation to the server, doing, for example :
asap-user-register --server vocms58.cern.ch --myproxy
In some case, when the ~/.glite/vomses is missing, it can give an error. The workaround is to create this file with the correct information :
mkdir ~/.glite cat $GLITE_LOCATION/etc/vomses/cms-voms.cern.ch.vo.ncm-vomsclient > ~/.glite/vomses
It can also fail for other reasons, like a library missing (libgridsite.so.1.1 for example). Always have a look in the crab.log to get the precise cause of the failure. This problem will disappear with CRAB client release 2.7.x and CRAB server 1.1.x because it does not use ASAP anymore.
- Doing a "crab -create", I get the error "Unable to connect to myproxy.cern.ch"
It means that CRAB has tried to create a proxy on myproxy.cern.ch, but it failed. The reason may be that our firewall (managed by the Computer Center team) has blocked the port. By default, myproxy-init takes the first port available in the range given by :
echo $GLOBUS_TCP_PORT_RANGE
The best solution is to change the minimum value of the range, doing, for example :
export GLOBUS_TCP_PORT_RANGE="20036,25000"
Notice that now, a script is doing it for you automatically, with a random minimum number, when you log on our user interfaces. So, you should not meet this problem anymore.
- Using CRAB, I got an error saying :
crab: <tt>voms-proxy-info -identity</tt> failed with exit code 256=0(signal)+1(status) crab: /C=BE/O=BEGRID/OU=IIHE/OU=ULB/CN=Marcel Dupont crab: WARNING: Unable to verify signature! Server certificate possibly not installed. Error: Cannot find certificate of AC issuer for vo cms
It probably means that the content of /etc/grid-security/vomsdir on your UI is not up to date. It can be solved by simply copying the content of this directory from an up to date UI.
- After having submitted with CRAB, my jobs stay a few minutes in "Submitting status", and after that, it says "Cannot submit".
Most of the times, it means that the CRAB server cannot find a compatible site. By "compatible site", we mean that CRAB looks for all the sites where your dataset is located, and for each it will check that the tag for your CMSSW release is published, and if the tag is not found in these sites, it ends with a "Cannot Submit". Another possible cause for this problem may be in your crab.cfg if you use white and/or black list. Perhaps you are to much restrictive. Try to submit without using back/white lists to see if the problem is still there. Also, never use se_white_list and ce_white_list together.
- Using CRAB271, the command "crab -create" gives an error message saying :
crab: Checking remote location crab: Problems trying remote dir check: Missing glite environment.
This bug and its origin are described here :
https://hypernews.cern.ch/HyperNews/CMS/get/crabDevelopment/1343/1.html
https://savannah.cern.ch/bugs/?65519
To avoid this, the workaround is simply to put :
check_user_remote_dir=0
in the [USER] section of your crab.cfg.
This problem is fixed in CRAB272.
- Using CRAB on SL5 UIs, I've got an error message saying :
Error - rank: Jdl mandatory attribute is missing
This can be obtained doing a "crab -match" or using CRAB in standalone mode.
This issue is coming from an outdated syntax in the file /opt/glite/etc/cms/glite_wms.conf.
Until it is corrected on all the machines, the workaround is to put the following line :
additional_jdl_parameters = rank =-other.GlueCEStateEstimatedResponseTime;
in the [GRID] section of your crab.cfg.
- My jobs finished with error code 60307 :
It means that the root files generated by your jobs could not be copied to the SE. If you use T2_BE_IIHE as storage element, then you should have also the following in your crab.cfg :
[GRID] virtual_organization = cms group = becms