Don't print pgrep advice if options not supported locally.

[hjoliver]

At start-up:

[INFO] *** listening on host01:43046 ***
[INFO]
[INFO] To view suite daemon contact information:
[INFO]  $ cylc get-suite-contact junk
[INFO]
[INFO] Other ways to see if the suite is still running:
[INFO]  $ cylc scan -n '\bjunk\b' host01
[INFO]  $ cylc ping -v --host=host01 junk
[INFO]  $ ssh host01 "pgrep -a -P 1 -fu $USER 'cylc-r.* \bjunk\b'" <--------------

The marked line is meant to print the parent PID and command line of a cylc-run or cylc-restart process for the suite, not those of the sub-process pool.

But it's been brought to my attention that some sites have a less capable version of pgrep:

pgrep: invalid option -- a
Usage: pgrep [-flvx] [-d DELIM] [-n|-o] [-P PPIDLIST] [-g PGRPLIST] [-s SIDLIST]
[-u EUIDLIST] [-U UIDLIST] [-G GIDLIST] [-t TERMLIST] [PATTERN]

[hjoliver]

We could either not print this line if the right pgrep is not present, or figure out how to get the same information without it.

[matthewrmshin]

Yes, a user has reported this to us some time ago as well.

These days, we actually have the PID of the suite running process at the point this message is printed out, so we can just use a ps command. See logic under:
https://github.com/cylc/cylc/blob/038e334bf017df2fb72d43b3152e96354ba9a974/lib/cylc/suite_srv_files_mgr.py#L137
which uses ps -opid,args PID to determine if the suite is still alive or not.

[hjoliver]

Good point, that would undoubtedly be simplest!

Also, we should not print the ssh HOST prefix. It should be obvious that you have to look for a process on the host that it is running on.

[matthewrmshin]

I guess we can do something like this.

[INFO]  $ ps -opid,args 123456  # on host1