"operation was canceled" from net not treated as context cancellation error

[julianbrost]

When a network operation from the net packages fails with net.errCanceled, the returned error err does not satisfy errors.Is(err, context.Canceled) which just resulted in a longer debugging session as this tore down errgroups because the error wasn't considered a context cancellation when it should be.

Unfortunately, there doesn't seem to be a really good way of accessing that error object, so to check it, you probably have to resort to comparing the error string.

This behavior was recently changed in Go, but the change isn't included in any released version: golang/go@85b5f86

[julianbrost]

This behavior was recently changed in Go, but the change isn't included in any released version: golang/go@85b5f86

Given that this is now included in released Go versions for a while (included from 1.19), this should not need any more work from our side. Our current release packages were built with a recent enough version (I checked only Icinga DB 1.1.0 ob Debian 10, that was built with Go 1.19.3, that version should be the same for all other distributions).

[yhabteab]

Actually, the issue still isn't fixed yet.

~/Workspace/go/src/icinga-notifications (trigger-time-based-escalations ✗) go version                                                          
go version go1.20.4 darwin/arm64

When using unix sockets it still crashes immediately.

2023-05-24T16:47:22.433+0200    INFO    icinga-notifications    Connecting to database at '/run/postgresql'
2023-05-24T16:47:22.434+0200    WARN    database        Can't connect to database. Retrying     {"error": "dial unix /run/postgresql/.s.PGSQL.5432: connect: no such file or directory"}
2023-05-24T16:47:22.434+0200    FATAL   icinga-notifications    cannot connect to database      {"error": "can't connect to database: can't retry: dial unix ....

But when using TCP it doesn't crash but keeps retrying:

2023-05-24T16:48:21.616+0200    INFO    icinga-notifications    Connecting to database at 'localhost:0'
2023-05-24T16:48:21.619+0200    WARN    database        Can't connect to database. Retrying     {"error": "dial tcp [::1]:5432: connect: connection refused"}
2023-05-24T16:53:21.620+0200    FATAL   icinga-notifications    cannot connect to database      {"error": "can't connect to database: can't retry: dial tcp [::1]:5432: connect: connection refused"

[julianbrost]

That's probably not a net.errCanceled (the socket isn't there isn't treated as a timeout), and there's already #543, feel free to test the corresponding PR #583.

[yhabteab]

NM! This is the wrong issue.