Linkerd in HA with 2 controllerReplicas #4618
Replies: 3 comments 4 replies
-
|
What API fails? |
Beta Was this translation helpful? Give feedback.
-
|
Its not api failure. Even with all required pods are up in one of the node(with second replicas are pending), unable to call the api(with linkerd stat). When checked with "linkerd check", same behavior and it halted at unscheduled pods Isnt it supposed to be working even with single replicas? |
Beta Was this translation helpful? Give feedback.
-
|
|
Beta Was this translation helpful? Give feedback.
-
|
Does that mean 3 replicas(on 3 nodes) required only for passing "linkerd check"? Linkerd api will be accessible and we can mesh application too with single replica right in case of two node failure? Also can we just include 3 replicas on 2 nodes too? |
Beta Was this translation helpful? Give feedback.
-
In HA mode, yes.
Yes.
Yes, but not in |
Beta Was this translation helpful? Give feedback.
-
|
I have just tested this, two different ways, api failed in both cases
See the error (in next post for readability ) when I try to "linkerd stat" or "linkerd inject" |
Beta Was this translation helpful? Give feedback.
-
|
]$ linkerd -n linkerd stat deploy $ kubectl get deploy siz -o yaml | linkerd inject - | kubectl apply -f - You can see all necessary pods are up in one node, other node, its pending. |
Beta Was this translation helpful? Give feedback.
-
Bug Report
As per the document https://linkerd.io/2/features/ha/, it looks we can limit the controller replicas to 2 for HA installation. I have deployed the same way in one node , it says "unschedulable pods" for the pods supposed to be scheduled on the second node. Second time, brought the second node up, old error is gone. However, shutting down one node fails the api, and it throws "error upgrading connection: error dialing backend: dial tcp :10250: i/o timeout"
What is the issue?
Linkerd API is inaccessible while running with 1 node with HA configuration(2 replicas)
How can it be reproduced?
Logs, error output, etc
stat pods
Cannot find Linkerd: error upgrading connection: error dialing backend: dial tcp NODEIP:10250: i/o timeout
Validate the install with: linkerd check
(If the output is long, please create a gist and
paste the link here.)
linkerd checkoutput× can initialize the client
error upgrading connection: error dialing backend: dial tcp NODEIP:10250: i/o timeout
see https://linkerd.io/checks/#l5d-existence-client for hints
Environment
Possible solution
Additional context
Beta Was this translation helpful? Give feedback.
All reactions