search

coreos / mantle

Mantle: Gluing Container Linux together
Apache License 2.0
95 stars 76 forks source link

Machine checks are flaky #1011

Open ajeddeloh opened 5 years ago

ajeddeloh commented 5 years ago

Kola tests sometimes fail the machine check because systemd is still in the starting state.

Specifically sometimes this check times out while stilling in the "starting" state: https://github.com/coreos/mantle/blob/master/platform/platform.go#L328

arithx commented 5 years ago

We can extend the timer (it's currently 300 seconds, 30 checks with one every 10 seconds) but at some interval it feels like there's a valid issue that the machine isn't finished starting.

ajeddeloh commented 5 years ago

I tried manholing in, but I kept getting weird terminal problems when I did (or it would fail on two machines at the same time and kola gets really confused when trying to manhole twice at the same time).