[MXS-4288] mxs1585 regression in 6.4 Created: 2022-09-10 Updated: 2022-11-02 Resolved: 2022-11-02 |
|
| Status: | Closed |
| Project: | MariaDB MaxScale |
| Component/s: | test |
| Affects Version/s: | 6.4.2 |
| Fix Version/s: | N/A |
| Type: | Task | Priority: | Major |
| Reporter: | markus makela | Assignee: | markus makela |
| Resolution: | Cannot Reproduce | Votes: | 0 |
| Labels: | None | ||
| Sprint: | MXS-SPRINT-166, MXS-SPRINT-167, MXS-SPRINT-168, MXS-SPRINT-169 |
| Description |
|
The mxs1585 test is unstable in 6.4 whereas in 2.5 it is much more stable and only fails due to VM related problems. The test appears to time out during a query which means this is possibly a real regression. |
| Comments |
| Comment by markus makela [ 2022-10-17 ] | ||||||||||||||||||||||||||||||||||||||||||
|
I managed to catch this once and it didn't appear to be caused by anything in MaxScale itself locking up. This would suggest that the reason is that a query is not routed and no errors are generated or a part of a result is not fully returned. | ||||||||||||||||||||||||||||||||||||||||||
| Comment by markus makela [ 2022-10-19 ] | ||||||||||||||||||||||||||||||||||||||||||
|
Now that the test framework collects the maxctrl create report output from tests that time out, I managed to catch the logs from two sessions for a failing test case. Both of them had been idle 308.6 seconds and were in the process of connecting to the galera cluster.
Based on the log output the handshake packet from server2 was either never sent or was not read by MaxScale. | ||||||||||||||||||||||||||||||||||||||||||
| Comment by markus makela [ 2022-11-02 ] | ||||||||||||||||||||||||||||||||||||||||||
|
The test no longer fails and the last 100 test runs of it seem to only fail due to overall slowness of the test system. Mostly it fails due to the cluster simply not responding to anything anymore. |