Application gets SIGTERM before next deployment is ready
yashko
PROOP

9 months ago

Hi team!

Today I see unexpected behaviour about deployments.

Before (few days ago):

I merge a PR, and new deployment is started. Old deployment is there, and application is working as expected.

After (around today):

I merge a PR, and new deployment is started. Old deployment receives SIGTERM and now my application have no active deployment while new deployment is still taking time, resulting in 2-3 minutes downtime.

I'm pretty sure I haven't changed anything in my settings.

Was there any change around this logic? How can I deploy like it was before?

Solved

3 Replies

brody
EMPLOYEE

9 months ago

Hello,

This is to be expected, unfortunately, your service has a volume, and since two deployments can not be mounted to the same volume at the same time, we must stop the previous deployment before starting the new deployment. This has always been the logic since we first introduced support for volumes.

Users have worked around this by offloading storage to a separate service such as MinIO so that their main application will not need a volume and thus is not subject to the mandatory downtime.

Best,
Brody

Status changed to Awaiting User Response Railway • 10 months ago

yashko
PROOP

9 months ago

Hi @brody! Okay I see. Will old deployment stay active if it doesn't rely on volume at time when second deployed started?

Status changed to Awaiting Railway Response Railway • 9 months ago

sarahkb125
EMPLOYEE

9 months ago

if the deployment doesn't point to the volume, it'll be unrelated.

+11

Status changed to Awaiting User Response Railway • 9 months ago

Status changed to Solved sarahkb125 • 9 months ago

Loading...