[MXS-4515] MaxScale leaks sessions if they are closed when writeq throttling is enabled Created: 2023-02-17 Updated: 2023-02-21 Resolved: 2023-02-20 |
|
| Status: | Closed |
| Project: | MariaDB MaxScale |
| Component/s: | Core |
| Affects Version/s: | 2.5.24, 6.4.5, 22.08.4 |
| Fix Version/s: | 2.5.25, 6.4.6, 22.08.5 |
| Type: | Bug | Priority: | Major |
| Reporter: | markus makela | Assignee: | markus makela |
| Resolution: | Fixed | Votes: | 0 |
| Labels: | None | ||
| Description |
|
The following Connector-J test case reproduces the problem. It only happens if writeq_high_water is lower than max_allowed_packet on the server. Due to the recent changes to the default values, this is unlikely to happen in older versions if the default values are used.
A workaround to this is to use the old default value for writeq_high_water. It won't eliminate the problem but it helps reduce the likelihood of it happening. Disabling the writeq throttling completely will prevent it but that can cause extreme memory usage. |
| Comments |
| Comment by markus makela [ 2023-02-18 ] | ||||||||||||||||||||||||||||||
|
Here's an example test program that reproduces the problem:
If the kernel is configured to use very small network write buffers or 2.5 is patched to tune them down (i.e. SO_SNDBUF set to 1024), the following assertion is hit in dcb.cc:
It turns out that this is a bug that was fixed for 22.08 but also needs to be fixed for 2.5. |