S
Supabase•2mo ago
jimdem

Big delays since yesterday evening

Hello, since yesterday evening i'm noticing big delays when i use my NextJs app in Vercel. I was using the same queries for weeks but now the delays are very big. Sometimes it is ok, but most of the times it is very slow. I'm seeing some issues like "Transaction API error: Transaction not found. Transaction ID is invalid, refers to an old closed transaction Prisma doesn't have information about anymore, or was obtained before disconnecting." or "Transaction API error: Unable to start a transaction in the given time." My code didn't change. Is there an issue in general?
54 Replies
Mr Joho
Mr Joho•2mo ago
Yes there is Experiencing same issues Same errors
paaulfff
paaulfff•2mo ago
any updates on this?
Mr Joho
Mr Joho•2mo ago
Status page shows all green and so far no response on urgent supabase support requests. Seems they are completely unaware of the issues so far.
paaulfff
paaulfff•2mo ago
the status page not updated is not a new thing it's the second time we have an issue with them, and the status page is not indicating any issues
Mr Joho
Mr Joho•2mo ago
For how long are you having the issues? For us roughly since yesterday 12 UTC. But much worse today
garyaustin
garyaustin•2mo ago
You are all getting transaction API error? The OP is using Prisma Are you all using Prisma? What regions?
paaulfff
paaulfff•2mo ago
yes, we are using prisma, and the same error
Mr Joho
Mr Joho•2mo ago
Migrations: aws-0-eu-central-1.pooler.supabase.com:5432
garyaustin
garyaustin•2mo ago
Are you using Supavisor pooling port?
Mr Joho
Mr Joho•2mo ago
Queries: aws-0-eu-central-1.pooler.supabase.com:6543/postgres?pgbouncer=true
paaulfff
paaulfff•2mo ago
we use pooling for database queries, but direct url for migrations and build time
garyaustin
garyaustin•2mo ago
What region and Supavisor port? If any of you are Pro you should to with pg_bouncer.
paaulfff
paaulfff•2mo ago
Central Eu - region
garyaustin
garyaustin•2mo ago
K. And you all have support requests in? I'll bump it as there are 3 of you.
paaulfff
paaulfff•2mo ago
supabase.com:6543/postgres?pgbouncer=true&connection_limit=60 yes, i did just made a support ticket
Mr Joho
Mr Joho•2mo ago
@garyaustin Thanks for bumping this. Using pg_bouncer and getting these issues: - Transaction API error... - Timed out fetching new connection from connection pool... - Unable to start transaction... And generally getting AVG response times of 60s that were previously <1s for weeks.
garyaustin
garyaustin•2mo ago
When I say use PG bouncer I mean if you are Pro pay for the IPv4 if you need it and use the dedicated PGbouncer port (not the flag prisma needs) you get as it gets you have the shared Supavisor pooler.
paaulfff
paaulfff•2mo ago
so, do we have to change someting or not? i dont really understand what's supposed to be the problem
garyaustin
garyaustin•2mo ago
My GUESS is there is a Supavisor issue limited to Central EU.
Mr Joho
Mr Joho•2mo ago
We are on Pro. Only using IPV4 for migrations. Might try for queries as well now.
garyaustin
garyaustin•2mo ago
Nothing you can do to fix a Supvisor outage if that is it. BUT if Pro Supabase recommends you use the direct Port 5432 and the PGbouncer installed with your instance now instead of Supavisor if you can.
paaulfff
paaulfff•2mo ago
so, do you work on that issue related to Supavisor? im gonna try with direct connection on 5432
garyaustin
garyaustin•2mo ago
There are no employees here if that is what you mean by do I... This is a user helping user forum.
paaulfff
paaulfff•2mo ago
but you are in the support team, right?
garyaustin
garyaustin•2mo ago
NO
Mr Joho
Mr Joho•2mo ago
Seemss to be more than just Central EU: https://github.com/orgs/supabase/discussions/38519
garyaustin
garyaustin•2mo ago
There are in general no employees here. Some of us are comped by Supabase to help out. I have no direct access to Supabase except thru a shared slack channel.
paaulfff
paaulfff•2mo ago
a ok, sorry, I saw your message related to the suport tickets
garyaustin
garyaustin•2mo ago
I'm asking that you have support tickets in. That is the only way they might know if not a "MASS" outage with others reporting to them. I'll add that discussion thanks. Just got up and had not looked thru Github yet.
v0idpwn
v0idpwn•2mo ago
Hi, I'm from the pooler team, would you please share the ticket id? I can then look closer
TrashTierDev
TrashTierDev•2mo ago
Hello, here is @paaulfff ticket number SU-243541
Mr Joho
Mr Joho•2mo ago
SU-243527
TrashTierDev
TrashTierDev•2mo ago
We have been having this issue since this morning, for the past 8-9 hours, some requests even take 4-5 minutes to complete
v0idpwn
v0idpwn•2mo ago
Thank you, Investigating
Mr Joho
Mr Joho•2mo ago
For us issues started at ~2pm yesterday and TTFB going from <1s to ~40s yesterday, to >1min today.
No description
v0idpwn
v0idpwn•2mo ago
Just to give an update, I identified a potential problem, will be trying a fix
TrashTierDev
TrashTierDev•2mo ago
Awesome, let us know !
v0idpwn
v0idpwn•2mo ago
Could you please take a look in your side to see if its faster for you?
TrashTierDev
TrashTierDev•2mo ago
It's still slow and I get error 500 on some requests it's a gamble, let's say I have 4 requests, it's either 3 are working, or just 2 are working, I just have to keep refreshsing until the stars decide to align and have all the queries working at once now i had instant loading of the data, but if i refresh again or do another request it takes a lot of time there is a new user on github with the exact same issue I have error : "\nInvalid prisma.client.findUnique() invocation:\n\n\nTimed out fetching a new connection from the connection pool. More info: http://pris.ly/d/connection-pool (Current connection pool timeout: 10, connection limit: 5)" Now it is a bit faster, but some requests keep having a 500 status
v0idpwn
v0idpwn•2mo ago
Yeah. I see some improvement in your metrics but its still not good.
Mr Joho
Mr Joho•2mo ago
These are some of the types of errors we are seeing.
No description
TrashTierDev
TrashTierDev•2mo ago
Same, we have several hundreds errors like that one
jonathanddipo
jonathanddipo•2mo ago
Everything looks good on our end at the moment. There have been no errors or timeouts for the last few minutes. We do not use prisma, we connect to supabase via node-postgres.
v0idpwn
v0idpwn•2mo ago
I'm restarting one instance now. You may see some impact (different errors?) but after that is finished, I believe timings should get better.
TrashTierDev
TrashTierDev•2mo ago
Okay, let us know when it's done, we are still having random completion times.
v0idpwn
v0idpwn•2mo ago
We will be opening an incident for this
TrashTierDev
TrashTierDev•2mo ago
Thank you, looking forward to your update
jimdem
jimdemOP•2mo ago
It was improved some minutes ago but i'm still getting errors like |Can't reach database server at aws-0-eu-central-1.pooler.supabase.com:6543|
Gabor
Gabor•2mo ago
Hello, we had the same issue. There were many timeout errors (> 15s) with Next.js serverless functions that query from Postgres at Supabase. The system is hosted at Vercel, we use Prisma, and the region is eu-central-1. We use the shared pgBouncer (port 6543). I could not see the error in the past hour.
TrashTierDev
TrashTierDev•2mo ago
I'm at under 6-7 seconds now compared to this morning where the app was not usable, maybe it will get improved once all is fixed.
v0idpwn
v0idpwn•2mo ago
It all looks stable for us for a while now.
Mr Joho
Mr Joho•2mo ago
Can confirm. Thank you @Void @garyaustin
TrashTierDev
TrashTierDev•2mo ago
Looks good on my side too, thanks guys ! Hope everything will be fine tomorrow 😄 Have a nice day/evening !

Did you find this page helpful?