greeting community. I have two instances of Prism...
# orm-help
a
greeting community. I have two instances of PrismaClient in my nodejs application (application lives in ECS contain in AWS if that matters). One client connects to reader db instance (postgres) and other connects to writer db instance. all of a sudden I get
"Timed out fetching a new connection from the connection pool"
which now does not disappear. there is no activity currently in the app but when try to do query get this error. Did anybody face such a problem and if yes what would be possible reason of the problem do get addressed? Any advice would be highly appreciated
r
@Alexei Snisarenko 👋 Could you try restarting the app on ECS or restart the DB and check if the error goes?
a
I’m sure it will go and intentionally not restarted to see the root cause
so as expected restarting the app got rid of the error. though the root cause in unclear. appreciate u replied back to my pretty vague request tho
🙌 1
r
We have seen a couple of messages like this but are unable to reproduce.
j
I have had a similar issue here @Alexei Snisarenko. If you figure out what caused this, would be curious to find out as well
It seems the auto-scaling for us influences the connection pools
a
autoscaling in ECS - will check as well. but not sure how single process should be affected. pool limit is 3 which is far from the available connections on the database. and there was only one process running with unavailable connections. I’ll share results here once find the root cause
j
Cool, id really appreciate that. Will do some analysis on this later this week as well
@Alexei Snisarenko Hey Alexei, was wondering if you and your team found anything out behind this? We’re currently doing some research and trying to debug the query engine and the mechanics behind the connection pool
a
Hi @Jonathan. From the moment I did restart the server the problem did not happen to reproduce. Aside from continuing watching over the server I so far do not know the exact circumstances the problem occurs and thus be able to get more information once the problem happens again
j
Gotcha, thanks for responding back. We will take care that when this happens, we can find a way to reproduce it
a
Sounds good