🐛 Bug Report: Aggregate ownership of large groups fail due to p-limit hanging after hitting max concurrent limit in plugin-orgs about backstage HOT 6 CLOSED

Yep that p-limit should not be reentrant; it'll deadlock. PR is in ^

from backstage.

@freben you rock. Thank you!

from backstage.

Hi @CiscoRob, I think this is a lot like this one? #22139

from backstage.

@awanlin I saw that issue when I searched for "aggregated", but wasn't seeing the same behavior. At least in terms of sheer URL request length or 431 status response we're not getting that. Didn't know if they had another issue or symptom of the same problem.

Each request in the network requests are all status 200 from our side. None fail. Just hits 12 requests and stops.

from backstage.

I'm wondering if it's related to p-lint. I see the package set to version 3.1.x and latest for the package is 5.x. Some updates related to async context handling have been made through version 4 and 5. Seems to get hung up after it gets past its maximum concurrent level of 10.

We're using bottleneck in some parts of our implementation for doing complex lookups for data exports and I've seen similar behavior from other schedulers.

from backstage.

@awanlin I pulled Backstage source, checked out v1.26.0-next.1, and imported the plugins-org into my local Backstage instance.

I confirmed / verified the "hanging" behavior was present in my local portal with the package directly referenced. I then removed the limiter() from the plugin-orgs and everything worked.

plugins/org/src/components/Cards/OwnershipCard/useGetEntities.ts

The limiter is definitely the issue and seems to just ... hang after any call past the 10 maximum concurrent count. Leaves the async action as if it's still "loading" but it never responds.

from backstage.