I’m getting a bunch of "Unhandled GraphQL subscrip...
# prisma-whats-new
a
I’m getting a bunch of "Unhandled GraphQL subscription errors” that seem to appear at random times even now when the service seems to be working fine. Any idea how to handle these? Is there a way to catch them?
n
@afgh thanks for the heads up we're about to fix it
@afgh could you check again?
Thanks for the heads up, we were able to identify the issue 🙂
a
My coworker tells me that she’s still been getting the error at random times since Friday 😞
n
huh, that's definitely unexpected. Is it something that you can reproduce consistently or does it pop up seemingly randomly from time to time?
a
So far it seems completely random. We haven’t seen one since Saturday though.
n
alright, please let me know right away if something comes up again, this would be very helpful 🙂
hey @afgh looks like a previous version of
subscriptions-transport-ws
had an issue that sounds like yours
could you update
subscriptions-transport-ws
to
0.5.4
? that's supposed to fix this problem
a
I’m on
0.5.5-alpha.0
n
huh interesting - could you try to go back to
0.5.4
?
a
Sure, but as I said we haven’t gotten the error since Saturday 🙂 I’ll let you know if I see it again with
0.5.4
.
👍 1
Just got it again.
Oh I'm still on
0.5.5-alpha.0
, sorry. Will downgrade now.
Same on
0.5.4
.
n
@afgh ah nice, Is it reproducible? what does the request look like in the network tab?
you can filter for WS only
a
Haven’t found a way to reproduce it, it seems to be random timeouts.
I’m not seeing anything weird in the network tab, there are finished and pending requests but I don’t see anything about errors.
n
could you please send the actual error message again?
a
n
and if you could get hold of such a faulty request in the network tab somehow that would be very helpful
at the moment it's hard to say if this is a problem on our end or with the websockets client
hey @alex_ye could you join the conversation here?
a
All right so I cleared the console and got the error again, there are no requests listed when filtering by WS.
n
Ah I see
a
Hi! My error message is a little bit different though "Subscription timed out - no response from server"
I’m using subscriptions-transport-ws 0.5.3
n
@alex_ye @afgh, looks like one of our subscription instances was overloaded. We're currently setting up infrastructure that better balances the load and automatically detects cases like these. Thanks for your support here!
👍 1
Ah, so your error @alex_ye might be fixed by upgrading ot 0.5.4. More information here: https://github.com/apollographql/subscriptions-transport-ws/pull/85
a
alright, I will try to upgrade. The thing is it didn’t happen until today
n
might also be a combination of the two issues 😛
a
@nilan the error is gone. But I still experience a significant delay of subscriptions. Can you please let us know, when the instance overload will be fixed?
n
hey @alex_ye could you pm me your project ID?
a
Still seeing the “unhandled graphql subscription error” messages @nilan 😔
Is it a load issue or could there be more to it?
n
there could be something else, can you send me the exact error message?
it sounds like a syntax error
we also rolled out an update to our subscriptions cluster that might have fixed this issue, could you have another look now?