Add log passthrough for lock-exec

[JoshStern]

Writes command output directly to std IO and no longer returns a buffer.

https://www.loom.com/share/fc29e6dce02e43059e495169468caf0e

[ejeun]

👀

just curious what you'll expect to happen with cron logs when this goes to prod. got any prediction?

[lingrino]

Just confirming I think this means that we're going to write output directly to stdout and return it in this function where it will be logged again by the lock-exec CLI?

I think that might be fine... What do you think about creating a new function like RunDirectOutput (not sure of a good name) that has your change here and then does not return the output. So instead of returning the output it just sends it directly to stdout and then returns only the command exit code.

[JoshStern]

I didn't want to change this function's interface so I set it up to do both.

This line logs it and adds the output as metadata to the log sent to dd. I was worried someone might have alerts or other things connected to that field.

I'm down to pick one or the other and drop that metadata if we think it's redundant though. Which would you prefer?

[JoshStern]

I could also update that log to avoid the double-print

[JoshStern]

@ejeun this should get us to a place where calling kubectl logs -f <job-pod> isn't empty if a job is hanging.

Haven't confirmed if these logs will make it to DD correctly but I was going to check it on staging/prod. I didn't see any special connections set up in lock-exec to connect to DD so I think it's at the stdout level, @lingrino can probably confirm.

[lingrino]

Yay! Thank you!

[lingrino]

I will merge this and tag a new version