Solution for JQM crash caused by pdls on a loaded server

2 Technical Update

HP Output Server 3.4.1

HP-UX specific Case

ulimit -a

Example output for ulimit -a command on Solaris:

$ ulimit -a

time (seconds) - unlimited

file (blocks) - unlimited

data (kbytes) - 96000

stack (kbytes) - 8192

coredump (blocks) - unlimited

nofiles (descriptors) - 256

vmemory (kbytes) - unlimited

In the above example, the data segment size is 96MB. So, if the pdls

operation on a heavily loaded JQM is executed and if the JQM process tries to

allocate memory more than the specified data segment stack size, the server

process will go to unknown state leading to failure of jobs and may dump core

Please check the help information of ulimit command information on each

operating system to change the data segment stack size.

Use indexed attributes while making pdls request

The JQM process buffers all data for the request and sends to the output when

you make a pdls request on JQM. It is suggested to avoid making pdls

request which could potentially return huge results.

For example: pdls -c job -a all jqm:

pdls -c job -a all -x"-job-status retained" jqm:

HP-UX specific Case

HP-UX uses different ARENA's for allocating memory for each thread. Due to

this reason, for every pdls request there could potentially be an increase in

the JQM memory usage if the request is served by different thread each time.

However, the increase would be limited to the number of threads used by the

JQM server process (Ex: 10 worker threads), after which JQM memory should

not increase or at least should not be as much as it did for previous requests.

This could be the reason for the crash only after 10 pdls and not during the

first one itself for the same number of jobs in it’s database.