Log file issue with parallel runs

[dpmatthews]

Users may have multiple UI server's running at once (e.g. one instance via Cylc Hub and another via cylc gui --new to test a new version). This can cause issues with the UI server log file.

To demonstrate the problem, run one instance of the UI server (cylc gui --new) and leave it running. It's log output appears in 01-uiserver.log. Then start up and shutdown another UI server - you'll find the log output from the original server is now going to 02-uiserver.log. Repeat this another 4 times and you'll find the log output from the original server is now going to a deleted file (depending on your filesystem, you may find it showing as a .nfs file).

I think a simple improvement would be to include the hostname in the log file name (the problem is most likely to occur when using multiple hosts with shared home directories). It may not be worth worrying about the case where you run multiple UI servers on the same host.

I suggest we look at this at the same time as #742.

[MetRonnie]

There is a downside to this with the use of floating pools of VMs where the hostname can change frequently - including the hostname in the log file name would leave behind lots of old log files. In #789 I am increasing the backupCount to 10 - is this good enough?

[MetRonnie]

Discussed today an idea to configure a different log file name on some uiserver hosts.

Unfortunately it doesn't seem possible to set c.CylcUIServer.logging_config['handlers']['file']['filename'] in jupyter_config.py.

Instead I suggest we use the value of the env var CYLC_UIS_LOG_FILENAME?

[MetRonnie]

Decision in today's meeting is to simply separate log files for hub-run UI servers and cylc gui-run UI servers (not making the file path easily configurable right now)