Random "Could not chdir to home directory".

[jiucenglou]

On PBS 18.1.2, many jobs can run without error but some randomly fail with Could not chdir to home directory in the error file. The job can run fine if resubmitted (till now). Could you help to suggest what could be the cause and how to fix ? Many thanks !

PS: There are a couple of occurrences of Could not chdir to home directory
in src/resmom/start_exec.c. I am still confused how to diagonalize the cause and fix ...

[zhenrong-wang]

Hi @jiucenglou

I am not an openPBS dev, but just went through the logic quickly. It seems that if you set the sandbox=PRIVATE, the openPBS would not change directory to the home directory, instead, it would change the directory to the job directory.

Maybe you can try this out.

[jiucenglou]

sandbox=PRIVATE

Thank you ! I will try sandbox=PRIVATE when the error shows up again and, since the error occurs randomly, see what happens afterwards in one or two months.

[jiucenglou]

sandbox=PRIVATE

I just saw the error again, and began to try sandbox=PRIVATE.

PS: (Fortunately) this time I was also running qstat -F json almost at the same time, and the command gave a corrupted json that jq refused to parse (other time jq can parse). I am wondering what could be the cause and how to diagnolize and workaround ? For example, is it possible to use qsub to ask pbs to "retry one more time" after "freezing/choking" ?