denoland / deploy_feedback

For reporting issues with Deno Deploy
https://deno.com/deploy
74 stars 5 forks source link

[Bug]: `TypeError: ops.op_set_keepalive is not a function` #382

Closed audionerd closed 1 year ago

audionerd commented 1 year ago

Problem description

My project has been running on Deno Deploy for the past 7 days untouched. It recently started throwing 500 errors, ops.op_set_keepalive is not a function. I’m using postgresjs@3.3.4, which is the latest version.

TypeError: ops.op_set_keepalive is not a function
    at TcpConn.setKeepAlive (ext:deno_net/01_net.js:162:16)
    at Object.setKeepAlive (https://deno.land/x/postgresjs@v3.3.4/polyfills.js:37:61)
    at connected (https://deno.land/x/postgresjs@v3.3.4/src/connection.js:365:51)
    at https://deno.land/x/postgresjs@v3.3.4/polyfills.js:150:32
    at Array.forEach (<anonymous>)
    at call (https://deno.land/x/postgresjs@v3.3.4/polyfills.js:150:18)
    at success (https://deno.land/x/postgresjs@v3.3.4/polyfills.js:103:11)
    at eventLoopTick (ext:core/01_core.js:166:11)
    at cachedError (https://deno.land/x/postgresjs@v3.3.4/src/query.js:171:23)
    at new Query (https://deno.land/x/postgresjs@v3.3.4/src/query.js:36:24)
    at sql (https://deno.land/x/postgresjs@v3.3.4/src/index.js:112:11)
    at readData (...)

Steps to reproduce

I was not able to reproduce locally. I think this may be a bug in Deno Deploy.

Expected behavior

The same code was running fine a few days ago.

Environment

No response

Possible solution

No response

Additional context

No response

bartlomieju commented 1 year ago

CC @lucacasonato I believe this is because of the change that causes disabled ops to be removed completely. You might want to add a stub op for that.

evelant commented 1 year ago

This is causing a total outage of my app (and I assume anyone else who uses postgresjs).

This appears to be a major breaking change applied silently. My code just suddenly started crashing without any changes due to this.

See issue in Supabase repo https://github.com/supabase/edge-runtime/issues/81

Anybody else that can be pinged to get this prioritized? @bartlomieju @lucacasonato

evelant commented 1 year ago

Looks like it's finally back up. I hope the Deno team will analyze how this happened and put steps in place to guard against it. A ~17 hour total outage was not a fun experience.

piscisaureus commented 1 year ago

We aim to retain backwards compatibility in Deno Deploy at all times, but unfortunately this change slipped through the cracks.

Our sincere apologies for the disruption.