https://www.prisma.io/ logo
Join Slack
Powered by
currently getting ``` Unhandled GraphQL subscripti...
# prisma-whats-new
w
currently getting
Copy code
Unhandled GraphQL subscription error [Error: Subscription timed out - no response from server
and
Copy code
VM25988:37 WebSocket connection to '<wss://subscriptions.graph.cool/v1/*******>' failed: Connection closed before receiving a handshake response
a
@wallslide Everything seems up: https://status.graph.cool/ Incidental or always?
w
It just seemed to have started happening.
I was testing some subscription-based code that was working fine, and it suddenly started occuring.
a
🙏🏻
Hope I didn't break anything with my large batch import tests 😄
w
I was thinking maybe its something to do w/ an out of date 
subscriptions-transport-ws
dependency
maybe you guys flipped a switch that disabled an older protocol or something
a
I saw something about 0.7.0 of that component
w
yeah, I'm on 
0.5.5-alpha.0
a
can you update?
w
yeah, about to. it's just that if it doesnt work, I dont know if its my setup or the server 🙂
a
well, at least your not using an outdated alpha version anymore 🙂
Did you see the post about the manual timeout setting on 0.7.0?
w
no
a
Maybe you can try that first
Manually override timeout setting in SubscriptionClient
Since you're getting a timeout...
w
looks like they fixed it in the latest 0.7.2 release
thanks for the heads up tho
a
Ah, didn't know, wasn't following the issue over there
gtk
w
it looks like you can perform all graphql functions via websocket now. Does graphcool support this? And would you recommend it?
a
They don't support it yet
w
okay. I couldn't see any documentation on setting up query batching over pure websocket transport anyways
😕
a
Haven't looked at it yet
w
still having subscription issues. hopefully the issue will solve itself 😅
e
On my side, I’m getting 
The connection to <wss://subscriptions.graph.cool/v1/>.... was interrupted while the page was loading.
in Firefox
w
glad its not just me
n
Hey guys I'm not aware of a problem let me have a look
l
I was seeing the issue noted here yesterday evening too. I've now updated to v0.7.2 and followed Soren's instructions for the timeout and everything's working again 👍
n
Ah good point! did you increase the timeout to 20000, @wallslide, @eliezedeck?
w
it doesn't look like the error is coming up anymore for me.
I was still getting it though after updating to 0.7.2, which has a default timeout of 30s
s
We had a temporary subscription issue this morning affecting a subset of projects. I am still investigating what caused it, but it seems to be unrelated to protocol versions and most certainly not related to data import performed by @agartha 🙂 A bit of background on how we currently operate the status page: We use a combination of cluodwatch alarms and external monitoring to detect service-wide issues. If any of these alarms are triggered, one of our engineers will be woken up. We then manually update the status page and start investigating the issue. Todays issue affected only a subset of projects and therefore didn’t trigger one of these alarms. The issue was ultimately resolved when the faulty server was automatically restarted. We are actively working on improving this process to automatically post minor incidents to the status page in a way that doesn’t introduce too much overhead for our engineering team. In addition to this we will be improving our failover mechanism to automatically re-route ws connections to healthy servers.
e
@nilan yes, I have timeout === 20000 ... still having this error right now
Copy code
The connection to <wss://subscriptions.graph.cool/v1/>... was interrupted while the page was loading.
@sorenbs @nilan I think you should give a bit more priority into the reliability of the service. It’s good to have nice and automated reporting, but I believe getting down to the source of the issue is more important; for me, server restart is not an actual solution. If it happened once, it sure is going to happen again unless it was resolved, reboot is just temporary. I would really love to run a Graph.cool-backed services, I cannot provide consistent services to my customers if my backend exhibits problems usually during the time where people in the USA are asleep and my customers are actively using the application. I know you are a business man, and I’m sure you understand the fear of having clients running away because of faulty services.
s
@eliezedeck Stability is our highest priority right now. Proper reporting is an important part of running a reliable service, so we are investing heavily in that as well as fixing bugs and making sure our systems are able to respond to issues in a fast and automated way.
Are you experiencing these errors right now? I think that might be a separate issue.
e
Yes, I am ... right now
Previously, it was displayed like every 20 seconds I presume. But now, it’s at the start of the application. Is that fine?
s
You really shouldn’t be getting errors like that at all. Are you able to receive subscription events when that happen?
e
This is what I’m trying to test now... because I had it setup only. I’ll let you know soon enough if this affects my subscriptions.
s
e
Nope, I get the same error on Chrome ... just with different error message
I’m sorry ... You’re right ... seems to be only in FF. I had the issue on both FF and chrome this morning thought. But since you restarted the server, it is now OK.
👍 1
Open in Slack
PreviousNext
Powered by