How set timeout in milliseconds?

[bzon]

RE: https://forum.nginx.org/read.php?2,29171,29181

For example, nginx.ingress.kubernetes.io/proxy-read-timeout: "500ms" is not valid.

[aledbf]

@bzon timeous < 1s are not supported
(the unit is hardcoded https://github.com/kubernetes/ingress-nginx/blob/master/rootfs/etc/nginx/template/nginx.tmpl#L1241-L1243)

[bzon]

Thanks for the clarification @aledbf.

I tried setting up the timeouts in millis using the server-snippet and it works. Will this override the controller's default timeouts that I saw configured within each nginx location?

[aledbf]

Will this override the controller's default timeouts that I saw configured within each nginx location?

The timeouts defined in the template are global. Using the server-snippet you are overriding that. All the locations in that server block will inherit your new value.

[bzon]

@aledbf is there a plan to support this in the future?

[aledbf]

Not really. Any change to the current approach could break existing definitions.
What you are doing, using the annotation is the way to define a custom time unit.

Edit: also, you should take the comment from the nginx thread about such low values

[bzon]

@aledbf

The timeouts defined in the template are global. Using the server-snippet you are overriding that. All the locations in that server block will inherit your new value.

Is this inheritance like an override? I ask this because I still see the controller's default timeouts in each location block along with the server-snippet configuration I added.

[aledbf]

@bzon I am sorry, the timeout https://github.com/kubernetes/ingress-nginx/blob/master/rootfs/etc/nginx/template/nginx.tmpl#L1241-L1243 is defined in the location, not the global scope.

Because of that, the only way to define such timeouts would be a custom template

[aledbf]

@bzon why do you want to setup low timeouts?

[bzon]

@aledbf our service average response time is 10-50 milliseconds and this is our SLO at 99.99%.

We just want to consider anything above 1 second to be a failure to prevent CPU build up.

I want to do the following:

• proxy-connect-timeout 100ms
• proxy-read-timeout 500ms
• proxy-send-timeout 500ms
• proxy-next-upstream-timeout 150ms

Edit: (additional settings)

• proxy-next-upstream-tries 2
• nginx.ingress.kubernetes.io/proxy-next-upstream: error timeout http_502 http_503 http_504 non_idempotent

[fejta-bot]

Issues go stale after 90d of inactivity.
Mark the issue as fresh with /remove-lifecycle stale.
Stale issues rot after an additional 30d of inactivity and eventually close.

If this issue is safe to close now please do so with /close.

Send feedback to sig-contributor-experience at kubernetes/community.
/lifecycle stale

[fejta-bot]

Stale issues rot after 30d of inactivity.
Mark the issue as fresh with /remove-lifecycle rotten.
Rotten issues close after an additional 30d of inactivity.

If this issue is safe to close now please do so with /close.

Send feedback to sig-contributor-experience at kubernetes/community.
/lifecycle rotten