a "maybe you forgot to reload" hint would do wonders

[leipzig]

I can't be alone here having tried to restart and debug a conf file and then suddenly remembering it has to be reloaded in this particular tool.

Just a thought:

/usr/local/bin/supervisorctl start production
production: ERROR (no such file). Maybe you forgot to reload?

[msabramo]

Yeah you're not alone. The thing that happens with us more commonly and is tricky is that someone updates settings for an existing app and someone forgets to do supervisorctl update so it's still running with old settings.

It would be neat if you could query for what settings are in use and/or the most recent last modified time (or a SHA-1?) for all of the conf files that were read at the time of the last read. The supervisorctl could warn when one of your conf files is newer than what is in effect and suggest that you do an update.

[msabramo]

I'd be willing to work on this if @mnaberez and @mcdonc gave some high-level direction.

[mnaberez]

It would be neat if you could query for what settings are in use

The reread command will reread the config and show differences to what is on disk without making any changes to the active configuration.

[msabramo]

Oh, that's what reread is. :-)

Dumb as it sounds I never quite looked into what reread does. I guess I could stand to RTFM a little bit more.

[msabramo]

I wonder if the reread functionality could be incorporated into status and the Web UI?

I feel like a lot of folks including myself mainly stick to start, stop, and status and/or the Web UI. They'd be more likely to notice a stale config in those places.

[mnaberez]

Originally posted in #1552:

I'd like to see if a process should be restarted when displaying the status.
It's easy to forget to reread the configuration file after modifying it. In such case, the user try to restart it and does not understand why the changes are not used.

Currently, the output is:

supervisor> status
demo                             RUNNING   pid 16371, uptime 0:02:43
demo2                            RUNNING   pid 16372, uptime 0:02:43

It could be something like:

supervisor> status
demo                             RUNNING  changed pid 16371, uptime 0:02:43
demo2                            RUNNING  -       pid 16372, uptime 0:02:43

Adding such info would break the backward compatibility if scripts parse the output.
To address this new problem, a configuration could be added to set the output of status in /etc/supervisor/supervisord.conf file. I think about two ways to implement it:

• 1. a boolean (called like changedstatus) to insert/ignore this data. By default, it must be disabled (for backward compatibility).
• 2. a list (called like statusline) to set up data displayed and their order. For example, with name,pid,status, a line will be like:

demo                             pid 16371, uptime 0:02:43  RUNNING

By default, it would be equivalent to statusline=name,status,pid (for backward compatibility).

IMO, Solution 2 is better because it's more versatile and allows to add new data in the future if needed.

Another way would be to add a new command to do that but the general usage of supervisorctl will be less clear so I think it's solution 2 is better than this one too.

What do you think about:

• the idea to add the changed data in status command?
• the way to implement backward compatibility?

I'm interested to make the PRs to implement it.