[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [HTCondor-users] best way to use cached data
- Date: Tue, 11 Dec 2012 09:36:36 -0600
- From: Todd Tannenbaum <tannenba@xxxxxxxxxxx>
- Subject: Re: [HTCondor-users] best way to use cached data
On 12/9/2012 11:36 PM, John Wong wrote:
I tried to run a script to read a file on the slave machine, but condor
couldn't read it even if it has full privilege. I think condor runs
every job in an isolated environment?
Overall, I think Ian Chesal's recent post gave good advice (as usual,
thanks Ian!) on how to accomplish what you want.
But I wanted to give a quick response on the "condor runs every job in
an isloated environment" question above. HTCondor may run the job as a
different user than the submit user, but I think you were already aware
of that. So as Ian implies, if Job A leaves files behind in /tmp that
are world-readable than a subsequent job can read them, *but* only if
you are not sandboxing directories like /tmp via the spiffy
MOUNT_UNDER_SCRATCH knob that appeared in HTCondor v7.7.x and works on
RHEL 5.x Linux and newer systems. By default, I don't think anything in
the filesystem gets sandboxed beyond having the current working
directory of the job in its own private space, but if you for instance have
MOUNT_UNDER_SCRATCH = /tmp
in your config file, then each job has its own private "/tmp" that is
destroyed when the job completes.