At startup, the processor kicks off an additional process to clean up the on-disk sym cache. Instead of running it on its own, it'd be better to switch to a process manager like Honcho to run it.

This reduces another complexity with the the processor architecture.

Making this block the GCP migration. I'll do this work before then.

Grabbing this to do soon.

I thought this would be pretty straight-forward.

1. write a bin/run_processor_honcho.sh that kicks off honcho using environment variables to configure things
2. write a processor/Procfile that runs /app/bin/run_processor.sh -- we already had this, so nothing changes here
3. change bin/entrypoint.sh to run honcho

I do that and suddenly none of the configman-style environment variables make it to the processor and it dies. I did some debugging:

1. bin/entrypoint.sh sees configman-style environment variables in the environment.
2. bin/entrypoint.sh runs /app/bin/run_processor_honcho.sh which sees configman-style environment variables in the environment
3. /app/bin/run_processor_honcho.sh runs honcho which runs /app/bin/run_processor.sh and the configman-style environment variables have been dropped from the environment

The honcho code has this code which restricts names to [A-Za-z_][A-Za-z_0-9]*, but I'm pretty sure that only runs on .env files and we're not using that.

https://github.com/nickstenning/honcho/blob/9eeb9b43e82b45c13de9999a0e65287cfbf384d6/honcho/environ.py#L56-L82

Then I traced how Honcho kicks off processes and what's going on is that it runs subprocess.Popen which on Linux runs popen_spawn which I think runs something like /bin/sh with the command. /bin/sh on Linux drops invalid variables including configman-style variables.

So... it looks like I have to fix configman stuff first.

I overhauled the configuration in Socorro so we're no longer using configman. That was in bug #1529342. As part of that, I changed the Socorro processor to run with Honcho.

All of that went out today in bug #1824987. Marking as FIXED.