Processor commands don't work correctly when clustered

CodeDrivenMitch / axon-open-admin

Axon Open Admin

MIT License

22 stars 6 forks source link

Processor commands don't work correctly when clustered #51

Closed oysteing closed 11 months ago

oysteing commented 1 year ago

Steps to reproduce:

Scale up backend to 2 (or more) nodes
Trigger Replay
Replay doesn't start. Some of the stop/start/restart commands return 204. Note that the backend instances share a DNS A name with multiple IP addresses.

I assume the 204 is a consequence of nodeId not matching the node the command is sent to.

CodeDrivenMitch commented 12 months ago

The 204 is indeed a consequence of that. Axon Open Admin only supports round-robin load balancers, as it relies on the browser to resolve the DNS. I will leave this issue open as a potential feature request. Perhaps querying the DNS manually, resolving it to multiple nodes and then registering both individually will work. But this needs research.

oysteing commented 11 months ago

There is actually a very concrete bug, introduced in 6465753. The 204 is simply ignored, although it is commented in the line below (https://github.com/CodeDrivenMitch/axon-open-admin/blob/master/frontend/src/components/processors/commands/CommandExecutor.ts#L82). I've submitted a pull request with a fix. The fix has been run successfully in production since February.