NodeJS process crash in MAA

[ignoramous]

2021-12-24T05:22:18.469 app[a6e3f003] maa [info] Main child exited with signal (with signal 'SIGINT', core dumped? false)
2021-12-24T05:22:18.470 app[a6e3f003] maa [info] Starting clean up.
2021-12-24T05:22:18.482 proxy[a6e3f003] maa [error] error.code=2000 error.message="problem connecting to app instance" 
2021-12-24T05:22:19.311 proxy[a6e3f003] maa [error] error.code=2 error.message="Internal problem" 
2021-12-24T05:22:20.222 proxy[a6e3f003] maa [error] error.code=2 error.message="Internal problem" 
2021-12-24T05:22:20.247 proxy[a6e3f003] maa [error] error.code=2 error.message="Internal problem"

Nothing more.

[amithm7]

This is not a crash, this is old instance shutting down. Happened during today's deployment 8 hours ago.

[amithm7]

Maybe strategy bluegreen or rolling should be used during deployment ?

I have found strategy bluegreen to be most stable during testing, but that also doubles instance count during deployment.

[ignoramous]

Yeah: Either a bluegreen or a canary deploy strat sounds okay to me. It looks like there may not be enough resources on shared-cpu instances to support bluegreen though.

[amithm7]

It looks like there may not be enough resources on shared-cpu instances to support bluegreen though.

It is a completely new instance (possibly at a new region). So, with bluegreen instances double temporarily and old ones shutdown together.

Currently, with default, only one instance is brought up and stabilized. All others are shutdown together. In observation, I couldn't connect to the app for about 30 - 60 seconds.

[ignoramous]

Okay, let's switch to bluegreen, even if it is more ambitious than a canary. Hopefully our testing in dev environment eliminates cause of such ambitious deployment strat.