Where's the webhook logs
I've a db webhook set up on a table on insert, lately I noticed it doesn't call my api, and I wonder why, is there anyway to troubleshoot this? perhaps I can find anything in logs?
GWebhooks are a postgres trigger function (postgres log if error) and a pg_net call. The net schema has two tables one of which is the actual call and the other is the http_reponse from your network request.
A_http_response
and
http_request_queue
?
the first has only a very few rows, the other is empty. I don't see any errors in the postgres log. I'm using the raw feature of Supabase Webhooks though I am aware that it is a trigger and pg_net under the hood. it has so far worked fine but only lately stopped unexpectedly, I thought it was my api, but my api is working when tested separately, and when a new row is inserted everything works in the db, except the db webhook isn't firing I believe.
and
http_request_queue
?
the first has only a very few rows, the other is empty. I don't see any errors in the postgres log. I'm using the raw feature of Supabase Webhooks though I am aware that it is a trigger and pg_net under the hood. it has so far worked fine but only lately stopped unexpectedly, I thought it was my api, but my api is working when tested separately, and when a new row is inserted everything works in the db, except the db webhook isn't firing I believe.
GIf you don't have entries in the request_queue and made very recent webhook calls then the trigger is not firing most likely.
GThe response is kept around longer and might show errors, but they would usually be response errors from the api.
GIf you do something that should generate a webhook call and go look at the request_queue table and don't see anything then your trigger is likely not firing OR erroring. But if erroring then should be in the Postgres log.
AI'm really surprised because it does work sometimes, and when it works I can see that my api has been called in my own api logs, but then sometimes my api isn't getting called. which made me suspect that the problem lies in the supabase webhook
GAs I said it is just a trigger that calls pg_net though. So one of those two parts would not be working. Trigger is all postgres so not likely a bug. pg_net does have alot more going on...
https://github.com/orgs/supabase/discussions/21023
https://github.com/orgs/supabase/discussions/21023
GitHub
This is a copy of a troubleshooting article on Supabase's docs site. It may be missing some details from the original. View the original article. NOTE: version 0.10.0 of pg*net is out. You shou...
GDid you set up the webhook with the UI or "code" your own trigger?
Ahow long does data stay in http_request_queue
when it worked once, I was watching the table but I didn't see any record.
when it worked once, I was watching the table but I didn't see any record.
AUi
GSorry, I don't know off the top of my head. It would be documented somewhere in the supabase/pg_net github. But it is not minutes I don't believe.
Athat's fine I'll keep trying, and debug, I hope I'll be able to find a solution
GOK that table is gone as soon as pg_net runs the request....
GSo it has a "loop" that checks that table, pulls off stuff to do then would delete the rows.
Ait has been working great for such a long time. I'm using supabase for almost two years, just lately started experiencing this particular issue
GHow fast are you hitting the table with changes?
Ado you mean how fast does the table get new records?
GIf that is what your webhook is on (inserts).
Ayes
Aso, not very fast
Aone new record every few minutes
GThen not sure. That is one thing that can be overwhelmed and there is a setting for that in pg_net (not sure about webhooks) but it is requests per second type thing.
Aon the same table I've triggers that also do make use of pg_net and they work always
GCould your API not be responding in time for any reason?
AI don't believe that is the reason, since my api gets in fact called from another trigger on the same table.
GSo you have multiple insert triggers both running pg_net very close together?
GThen why are you using the webhook versus just your triggers?
Aone I setup myself, calls a different endpoint, than the one the webhook calls
GNot that I know of a reason that would not work.
AI've a notify endpoint, which it was really easy to use with the webhook, the other endpoint I had to use a trigger because there's some logic that goes inside the trigger
GIt should look like trigger1 pg_net return, trigger2 pg_net return and pg_net would not know the difference.
GWhat are your timeouts set for in your code versus the webhook?
G
GThe trouble shooting link I gave is likely not much use if another pg_net call is working at same time.
GAny errors in the http_response table rows?
Athis is in the code for my first trigger (the one I setup manually) which is working.
Athere aren't any.
A
GYou might increase that.
I think default pg_net is 5000.
I think default pg_net is 5000.
GWhich is what your other trigger is running.
GThat says your API has 1 sec from the time the request leaves Supabase to respond back to Supabase including latency.
AI don't think I've manually changed it to 1000, is it possible that it was the default in the UI a while back?
GNo idea what the default was.
Let me see what it is now.
Let me see what it is now.
Aoh. I'll try increasing that for sure then
GIt is 5000 now.
GI think that would fit your issue. If you were right on the edge and things slowed down a bit.
AYou're right. I'm seeing that it now hitting my api versus before it only hit once at such random times.
GI would think though some rows in response table would have error, but more detail than I've ever looked into.
Awhen I changed that to 5000
AI'm not seeing errors in the tables we talked about nor in the other logs of supabase. But that it now works hopefully it will continue to work, only time would tell
