upforwd test failure
When CI is running low, the upforwd test fails. Previously this could cause an assertion failure, which has now been fixed by #4064 (closed), but we still need to address the failure.
The failure is described at #4064 (comment 372499). It usually looks like this:
I:upforwd:checking DNSTAP logging of UPDATE forwarded update replies (26)
I:upforwd:UQ=4 UR=1
I:upforwd:failed
...
I:upforwd:checking DNSTAP logging of UPDATE forwarded update replies (28)
I:upforwd:UQ=3 UR=1
I:upforwd:failed
As I said above, I suspect this is because the test platform is slow: some update forwarding requests are timing out before replies are sent, and so we count more requests (UQ) than replies (UR) in the DNSTAP stats.
I'd love to find out why these CI jobs are so draggy. It shouldn't be taking 30-40 minutes for a system test job to run. It's possible there's some environmental issue, and if we fixed it the test failures would go away. That said, however, I think we can fix the test so it's more resilient.