supercronic logs "error running command: waitid: no child processes" regulary since v0.2.31

[ceesgeene]

Since https://github.com/aptible/supercronic/releases/tag/v0.2.31 we regulary see superconic logging errors with the following message: "error running command: waitid: no child processes"

In that release ramr/go-reaper was added to automatically reap zombie processes (see 8e8973c ).

For that library there already exists a similar issue: ramr/go-reaper#11

One of the suggestions (ramr/go-reaper#11 (reference)) is to use a more verbose chunk of code so that the reaper runs inside a different process: https://github.com/ramr/go-reaper?tab=readme-ov-file#into-the-woods . Not sure if this is applicable to supercronic

[UserNotFound]

Hi @ceesgeene, can you clarify if there is any impact beyond logging the error?

Pinging @qianlongzt in case you're able to assist with a triage or fix.

[qianlongzt]

The impact is just random (not 100%) waitid errors. The CronsFailCounter may incorrect

Perhaps the fix is to run the reaper first and then start supercronic as its child?

https://github.com/aptible/supercronic/blob/master/cron/cron.go#L256
https://github.com/aptible/supercronic/blob/master/cron/cron.go#L119

[jonasgeiler]

I am having the same problem. Funnily enough it seems like using --init when running the container mitigates this problem? Or maybe I just never witnessed it when using --init. As @qianlongzt said, it is very random.