Add synctl warning when no process is stopped and fix restart

[romainbou]

As suggested in #5166, I added a warning when no process is killed for synctl stop and synctl restart
For synctl restart if no process is killed don't attempt to start anyway.

Example stopping synapse twice:

> synctl stop 
stopped synapse.app.homeserver

> synctl stop
No running worker found. (from '~/synapse/homeserver.pid')
The process might be managed by another controller (e.g. systemd)

> synctl restart
No running worker found. (from '~/synapse/homeserver.pid')
The process might be managed by another controller (e.g. systemd)

[clokep]

Thanks for your contribution! Sorry for the delay in reviewing this! Looks pretty good overall, but I left a few comments.

[clokep]

@romainbou Can you fix the lint issues? You can run them locally with: ./scripts-dev/lint.sh

[clokep]

@romainbou Do you have plans to take a look back at this PR? It'd be great to finish it up!

[clokep]

@romainbou Can you merge develop into this branch? There's been some changes to CI and such since the PR was first opened!

[romainbou]

@clokep cool thanks yes indeed. After merging develop again, now all checks are passing!

[clokep]

Looks good! Thanks for making those changes! 👍

[richvdh]

this seems to have had the side-effect of changing the behaviour of synctl restart so that it doesn't start synapse if it wasn't running previously. Was this a deliberate change? IMHO it's somewhat unintuitive and inconsistent with the behaviour of other process managers.

[richvdh]

it seems that it was a deliberate change, and indeed that was somewhat suggested by #6994.

On reflection, I feel like the downsides of that change outweigh the upsides (specifically: it's a surprise to anyone who has ever used synctl or any other process manager that synctl restart does not start synapse if one was not previously running). Accordingly, I have reverted that part of the change in #7624. The new warning text is retained.

[richvdh]

I should also mention: I don't think that changing the behaviour of synctl restart in this way brought any particular advantages. (The goal was to help people who blindly assumed that the stop part of synctl restart had worked: whether or not we go on to do a start after a failed stop has no bearing on that.)

An even better fix here is to sort out #4641, but for now the warning text is an improvement.