Unactionable Gitaly "fatal: the remote end hung up unexpectedly" errors may trigger our SLI alerts
Summary
In production#7002 (closed) we had a single user/repo triggering errors with cmd wait: exit status 128, stderr: "fatal: the remote end hung up unexpectedly\n"
on RPC SSHUploadPackWithSidechannel
at a sufficiently high rate to trigger our single-node error rate alerting. We haven't been able to establish the root cause of these errors (the user was contacted by support but so far we've received no reply). If we settle on this being a user-originated error, we should look for a way to exclude it from our error metrics.
Related Incident(s)
Originating issue(s)
Edited by Steve Xuereb