Skip to content

GATEWAY-object randomly set Programmed False, Loadbalancer IP still exist on service that handle GATEWAY-object. #5715

Open
@nwe01

Description

@nwe01

Description:

What issue is being seen? Describe what should be happening instead of
the bug, for example: The expected value isn't returned, etc.
After Gateway is deployed we get External/Internal IP-address from GKE-loadbalancer.

During daily operations the gateway randomly looses connection to the underlying service, and get status PROGRAMMED: False, but the service still have the IP-address assigned"

We initially suspected that this could have something to do with the HPA and autoscaling, but we have ensured that we are not setting any static replicas and that we are only setting minReplicas and maxReplicas with the HPA.

we have also done some load testing to ensure that the gateways are not set to "PROGRAMMED: False" during scaling operations.

The Message from Gateway is when it set to False is this..

"No addresses have been assigned to the Gateway"
and we are hitting this function.

https://github.com/envoyproxy/gateway/blob/main/internal/gatewayapi/status/gateway.go#L143

We have test both version 1.30 and version 1.32 and still hit the issue..

$ k get gateway -n gw
NAME CLASS ADDRESS PROGRAMMED AGE
gw-public eg-public 35.XX.XXX.XX False 8d

Sometimes GATEWAY is set to FALSE.
$ k get gateway -n gw
NAME CLASS ADDRESS PROGRAMMED AGE
gw-public eg-public False 8d

but the service for gateway gw-public/internal still has IP Assigned
$ k get svc -n gw
gw-public-cc293c26 LoadBalancer 10.145.6.0 35.XX.XXX.XX 80:31781/TCP,443:30118/TCP 8d

After around 15 - 60 minutes Gateway are coming back with PROGRAMMED True. and we can reach our site/url.

Metadata

Metadata

Assignees

No one assigned

    Labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions