UncaughtException from some timeout makes CI jobs fail

[last-genius]

Not sure which side is at fault for the timeout:

Restoring opam cache
  node:internal/process/promises:391
      triggerUncaughtException(err, true /* fromPromise */);
      ^
  
  AggregateError [ETIMEDOUT]: 
      at internalConnectMultiple (node:net:1118:18)
      at internalConnectMultiple (node:net:1186:5)
      at Timeout.internalConnectMultipleTimeout (node:net:1712:5)
      at listOnTimeout (node:internal/timers:583:11)
      at process.processTimers (node:internal/timers:519:7) {
    code: 'ETIMEDOUT',
    [errors]: [
      Error: connect ETIMEDOUT 8.43.85.97:443
          at createConnectionError (node:net:1648:14)
          at Timeout.internalConnectMultipleTimeout (node:net:1707:38)
          at listOnTimeout (node:internal/timers:583:11)
          at process.processTimers (node:internal/timers:519:7) {
        errno: -110,
        code: 'ETIMEDOUT',
        syscall: 'connect',
        address: '8.43.85.97',
        port: 443
      },
      Error: connect ENETUNREACH 2620:52:3:1:0:246e:9693:128c:443 - Local (:::0)
          at internalConnectMultiple (node:net:1182:16)
          at Timeout.internalConnectMultipleTimeout (node:net:1712:5)
          at listOnTimeout (node:internal/timers:583:11)
          at process.processTimers (node:internal/timers:519:7) {
        errno: -101,
        code: 'ENETUNREACH',
        syscall: 'connect',
        address: '2620:52:3:1:0:246e:9693:128c',
        port: 443
      }
    ]
  }

For example, see this CI job: https://github.com/xapi-project/xen-api/actions/runs/13852796742/job/38763073410?pr=6359

[ejrgilbert]

Same here (different repository): https://github.com/ejrgilbert/whamm/actions/runs/13858678635/job/38781385263

[mt-caret]

Observing the same issue here: https://github.com/mt-caret/duckdb-ocaml/actions/runs/13859018593/job/38782518500.

[ejrgilbert]

I think there's some spottiness for a service that this depends on. I've gotten CI to work if I just rerun it a couple times.

[smorimoto]

I believe this is caused by a failure of the GitHub Actions cache server. I implemented this action to allow OCaml to be set up without a cache server if for some reason the cache server is not available, but apparently the upstream package didn't handle the timeout correctly. We can hack it, but let me check the status upstream first 🤷‍♂️

[mbarbin]

+1 I started to see similar errors at a pretty high rate in my CIs over the last $N days (perhaps N = 2, 3?).

[mbarbin]

@smorimoto just wanted to say thank you as I re-run the CIs jobs that had this error and all of them succeeded now. I am guessing that if the jobs config specifies ocaml/setup-ocaml@v3 it is enough to have picked up automatically the latest release you created with the fix (is that right?). Thanks a lot for the fix!

[smorimoto]

Yes! It’s totally enough!

[mbarbin]

@smorimoto Some other jobs is now failing with an error that looks possibly related or similar:

Restoring opam cache
  node:internal/process/promises:391
      triggerUncaughtException(err, true /* fromPromise */);
      ^
  
  AggregateError [ETIMEDOUT]: 
      at internalConnectMultiple (node:net:1122:18)
      at internalConnectMultiple (node:net:1190:5)
      at Timeout.internalConnectMultipleTimeout (node:net:1716:5)
      at listOnTimeout (node:internal/timers:583:11)
      at process.processTimers (node:internal/timers:519:7) {
    code: 'ETIMEDOUT',
    [errors]: [
      Error: connect ETIMEDOUT 8.43.85.97:443
          at createConnectionError (node:net:1652:14)
          at Timeout.internalConnectMultipleTimeout (node:net:1711:38)
          at listOnTimeout (node:internal/timers:583:11)
          at process.processTimers (node:internal/timers:519:7) {
        errno: -110,
        code: 'ETIMEDOUT',
        syscall: 'connect',
        address: '8.43.85.97',
        port: 443
      },
      Error: connect ENETUNREACH 2620:52:3:1:0:246e:9693:128c:443 - Local (:::0)
          at internalConnectMultiple (node:net:1186:16)
          at Timeout.internalConnectMultipleTimeout (node:net:1716:5)
          at listOnTimeout (node:internal/timers:583:11)
          at process.processTimers (node:internal/timers:519:7) {
        errno: -101,
        code: 'ENETUNREACH',
        syscall: 'connect',
        address: '2620:52:3:1:0:246e:9693:128c',
        port: 443
      }
    ]
  }
  
  Node.js v20.19.0

This is taken from this job which I believe is using a version with the patch.

[smorimoto]

It's weird that the issue is still happening, let me debug it... 😮‍💨