An intermittent error that we observe reasonably frequently is an I/O
timeout that occurs while start/stopping a service hundreds of times,
and an error is returned as a Postgres connection was being established:
Here, try to address the problem by special casing for an "i/o timeout"
error if one comes back during a stress test. The approach is a little
on the hacky side, but given these tests are a little extravagant anyway
in the amount of churning they produce on purpose, I think it's probably
okay.
An intermittent error that we observe reasonably frequently is an I/O timeout that occurs while start/stopping a service hundreds of times, and an error is returned as a Postgres connection was being established:
Example failing run here:
https://github.com/riverqueue/river/actions/runs/9771997514/job/26975810964?pr=416
Here, try to address the problem by special casing for an "i/o timeout" error if one comes back during a stress test. The approach is a little on the hacky side, but given these tests are a little extravagant anyway in the amount of churning they produce on purpose, I think it's probably okay.