Getting realtime errors after few minutes of idle on web

[cRaZy92]

Hello,
I am getting request errors while I am not doing anything on my web that is using Pocketbase.
There are 2 real-time connections active and after a while, I get errors like this:

There is no further information, when I click on the error, it redirects me to a specific request:

It's not really broken but it gets old with that error in the console like 100s times one after another.
I am not sure what else I can provide, please let me know.

Thanks for your help!

[ganigeorgiev]

Hm, maybe there is a fly.io limit of how long a http connection could remain open. I haven't really tested the realtime api with Fly.io. Will try to find some time later today or tomorrow and will see if I can reproduce it.

In any case, there should be an auto reconnect in place, so even if there is a limit, you can ignore it and the SDK will try to keep the realtime connection alive.

[dschissler]

@ganigeorgiev @sgtpep What about if Pocketbase could handle the lastEventId? It would need some sort of internal caching though.

[ganigeorgiev]

@dschissler It is not possible to use the lastEventId because at the moment we don't store the messages anywhere but I have it in my todo and will consider it in the future (most likely together with allowing passing filter expr in addition to the subscription topic).

[dschissler]

@ganigeorgiev The magic there would be if subscribing filters are the same as querying filters. Also allowing it to somehow include the results of a query on connect based upon some variables / parameters that were sent and with lastEventId. Of course that is a lot of work and probably with a lot of edge cases. It seems pretty great though.

[sgtpep]

I'm relatively new to realtime stuff. One of the problems that I'm trying to solve with the current implementation of realtime in PocketBase is making sure no record updates are lost on reconnections due to timeouts and going offline.

I vaguely see how lastEventId could be useful for such scenarios, but until that's added to PocketBase, I found a workaround: I subscribe to PB_CONNECT: pb.realtime.addEventListener('PB_CONNECT', ...) and re-fetch the whole record(s) every time it fires to re-sync data.

Definitely, would be great to have a more straightforward way of doing that (possibly a more high-level one).

[MaCleodWalker]

@sgtpep Im trying to solve a similar issue could you provide a little more detail on what you did? Thank you.

[raizdev]

Im facing exact the same issue. We are behind docker swarm and our own nginx configuration (its pretty standard tbh), seems that the connection lost after 2 minutes.

[ganigeorgiev]

@raizdev Please open a new discussion with more details about your use case and deployment setup.
This discussion is related to a default Fly idle timeout and your problem seems to be caused by something totally else.
For example nginx config you can refer to https://pocketbase.io/docs/going-to-production/#using-reverse-proxy (notice the proxy_read_timeout directive).

[raizdev]

You are correct. I totally missed that part. I’m sorry it’s fixed