[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [HTCondor-users] Start HTCondor service under valgrind.

On 2/1/22 12:00, Sergey A. Komissarov via HTCondor-users wrote:

I'm trying to envistigate why condor_schedd crashed with valgrind but HTCondor daemon just refuses to start. Does it need some options to work correctly under valgrind?
HTCondor 9.0.9 is started inside Ubuntu docker container. When I replaced usual start command (/usr/sbin/condor_master) with valgrind HTCondor process just dies after 1.5 minutes.

Condor start command: valgrind --tool=memcheck --verbose --trace-children=yes --child-silent-after-fork=yes --trace-children-skip=/usr/bin/*,/bin/* --time-stamp=yes /usr/sbin/condor_master
Valgrind complete log:

We run condor under valgrind routinely, but usually not in a container, on raw hardware. There is a known problem with running condor under valgrind, which condor tries to work around. For the schedd to start a shadow when running under valgrind, the config knob

USE_CLONE_TO_CREATE_PROCESSES must be set to false. Condor itself tries to check that it is under valgrind, and if it detects this, automatically sets the knob to false, and logs something like this to the SchedLog:

Looks like we are under valgrind, forcing USE_CLONE_TO_CREATE_PROCESSES to FALSE.

Otherwise, is there anything interesting in the SchedLog when it exits?