search

waku-org / nwaku-compose

Deployment docker-compose files to deploy an nwaku node
12 stars 16 forks source link

Revert "Remove RPC (#65)" #69

Closed gabrielmer closed 5 months ago

gabrielmer commented 5 months ago

This reverts commit 94b40059c938fdcdbc9f5246a30361aad5db1f23.

Since that commit, the Grafana dashboard looks the following way

image

Might be related to removing the RPC from nwaku-compose without having the RPC deprecation commit in the release?

Reverting the commit until finding the root cause.

alrevuelta commented 5 months ago

Wait, we can't "Reverting the commit until finding the root cause.". First we find the rootcause and then we revert whatever is needed. Otherwise there is no valid reason to revert.

Said that, I don't think it's related.

Might be related to removing the RPC from nwaku-compose without having the RPC deprecation commit in the release?

Note that the reverted PR just disables the (afaik) unused rpc. Dont think its related to the rpc deprecation release.

gabrielmer commented 5 months ago

Wait, we can't "Reverting the commit until finding the root cause.". First we find the rootcause and then we revert whatever is needed. Otherwise there is no valid reason to revert.

So users were showing that they weren't able to run a node properly in a way that hadn't happened before. I ran the commit merged today and same thing happened to me. Went one commit behind and it worked.

If indeed the last commit let the node in an unusable state for the users, then best thing IMO is to revert and figure out the root cause while it's still usable for users. Don't know if it makes sense, that was my line of thinking.

I'm also not sure how this change would cause this - I'm now thinking it's not related because it is also happening sometimes with the reverted commit. But when I tested it and reverted it, the previous commit worked and today's didn't.

Apologies if it just messed things up more 😶

gabrielmer commented 5 months ago

It seems that it's something related to discv5 not working properly. The times that I saw the visualization Discv5 (Seen Nodes) get populated at startup, everything eventually worked. When at startup that visualization doesn't get populated, it gets stuck at that state.

There doesn't seem to be a consistent behavior, sometimes it works and sometimes it doesn't - which is alarming.

Edit: now I got a run in which Discv5 nodes got populated but the rest not, so that theory might also be discarded