HA conflicts with git storage #3676
Replies: 2 comments
-
I'm seeing this exact same issue. We're scaling back to be non-HA for the moment, but would love to know the current state of understanding around this. |
Beta Was this translation helpful? Give feedback.
-
We are also seeing git conflicts with HA mode enabled. We run two replicas of wikijs as Kubernetes pods, round-robin load-balanced by a regular Kubernetes service. With multiple replicas, there are often merge conflicts between the two. My best guess is that either two people are editing the same page simultaneously (how is this usually handled in the non-HA case, I'm not sure?) using different replicas; or the same person makes a change to a page, saves it, edits the page again (but this is served by a different replica) and saves it - all within the five minute git push window. Therefore the replica which pushes first causes a merge conflict on the other replica:
Thanks! |
Beta Was this translation helpful? Give feedback.
-
Describe the bug
I've 2 EC2 instances running WIkiJS, installed through Ansible with the same settings, postgres database over RDS. Behind HAProxies with Consul based service discovery. Git storage is enabled for some integration.
ha: true
and wikijs's2.3.77
.There are a bunch of error message when a change is made, see the screenshots below. Some of the error message appear in systemd's journal. And I can see a
My guess is, one instance is committing, pushing, the other get the change through the HA mechanism, try to commit & push too, & boom. Or something alike.
To Reproduce
Steps to reproduce the behavior:
Expected behavior
Users should be unaware of HA, no such message being displayed
Screenshots
Host Info (please complete the following information):
Beta Was this translation helpful? Give feedback.
All reactions