[HTCondor-users] Job's DiskUsage resolves symlinks?

[Oliver Freyermuth <freyermuth@xxxxxxxxxxxxxxxxxx>]

Dear HTCondor experts,

I'm not sure if this is a bug or feature: We are currently not enforcing DiskUsage_RAW to stay below the Request_Disk, but have started to monitor our user jobs for cases when they mismatch their request,
preparing for a future enforcement.
One of our users has found a huge mismatch: He only used a few GB in his execute directory, but he also added some symbolic links to data stored outside of the execute directory
(on our shared file system), and DiskUsage_RAW rose to a disk usage higher by several factors.

It seems the symbolic links are followed and the disk usage reported via ATTR_DISK_USAGE includes the size of the files behind the links.
Of course, these files don't actually take up space in the execute directory, though.

Is this on purpose?
It is natural to follow symbolic links when determining the size of directories for HTCondor file transfer, since the data will actually be transferred,
but for the space "consumed" by the execute directory, this does not match my expectation.

Checking the sources[0], I find the actual size check is done via stat() if lstat() shows the file is a symbolic link, so links are explicitly dereferenced
(and this code is of course used for any size checking tasks).

Cheers,
	Oliver

[0] https://github.com/htcondor/htcondor/blob/789360c371546f9c86904f1c8439efe7cef8c409/src/condor_utils/stat_info.cpp#L141-L148

--
Oliver Freyermuth
UniversitÃt Bonn
Physikalisches Institut, Raum 1.047
NuÃallee 12
53115 Bonn
--
Tel.: +49 228 73 2367
Fax:  +49 228 73 7869
--

Attachment: smime.p7s
Description: S/MIME Cryptographic Signature