a toaster run leaves orphaned processes

[tbrowder]

I’ve looked at Proc::Q and Toaster and I (a concurrency noob) can’t see any obvious leak. However, child processes are being orphaned somehow.

As a temp work around, I might try keeping a list of the proc pids and, after all modules are handled, try to ensure all pids are reaped.

UPDATE

I have just had a successful toaster run with no orphans and I think the culprit in this saga may be me:

Not yet stated is how I’ve been actually executing a run. Since the job is long-running I have been using nohup. After reading about possible problems with stdin I started experimenting with this method:

nohup time perl6 bin/toaster-perl6 > toaster.out 2> toaster.err < /dev/null &

but that apparently blocks use of stdin which the job needs, so I next dropped the stdin redirect:

nohup time perl6 bin/toaster-perl6 > toaster.out 2> toaster.err &

Then, still seeing orphaned processes, I started experimenting with various tweaks to Proc::Q trying to ensure no child processes were being left alive. After many runs, I started looking more closely at the orphan situation (using top and ps ajfx) and decided that nohup may be the root of the problem.

I then rediscovered the at utility, which I haven’t used for many years since leaving a heavy multi-user environment early in my second career. Bingo, no orphans on the first run! Now I have untweaked Proc::Q back to its original form and am trying another run:

I created a one-line script, run-toast.sh, to hold my desired command which is the same as the last nohup line above but without the nohup and the ampersand. Then I executed the file:

at -f run-toast.sh now

and it’s off and running in the default at queue.