[MXS-1833] Connection timeout when creating CONNECT tables for maxinfo in Galera cluster with SSL set up Created: 2018-04-26 Updated: 2020-08-25 Resolved: 2018-04-27 |
|
| Status: | Closed |
| Project: | MariaDB MaxScale |
| Component/s: | maxinfo |
| Affects Version/s: | 2.2.4 |
| Fix Version/s: | 2.2.6 |
| Type: | Bug | Priority: | Critical |
| Reporter: | Valerii Kravchuk | Assignee: | markus makela |
| Resolution: | Fixed | Votes: | 0 |
| Labels: | galera, regression, ssl | ||
| Description |
|
It seems some change in MaxScale 2.2.x or maxinfo there comparing to older 2.1.x versions leads to the following problem while creating CONNECT table to monitor MaxScale via maxinfo service in Galera clusters with at least 2 nodes:
It seems with only one node there is no problem. Related settings from my testing environment are: 1. maxscale.cnf:
2. Two nodes (10.2.x) have the following configuration files:
3. To reproduce, install maxscale and start it as a service. Initialize node1, start it as a new cluster, then start another node and make sure they formed the cluster:
Connect to maxinfo at port 9003 and make sure SHOW commands works as expected. Then connect to maxscale router at port 4008 as -umyuser -pmypwd (the user is configured for SSL and certificates are created based on this blog post, https://mariadb.com/resources/blog/mariadb-maxscale-21-and-ssl-certificates), load CONNECT engine plugin, and then run this script:
You should see error message like this for some tables: ERROR 1159 (08S01): Got timeout reading communication packets I've seen cases when one of CREATEs just hanged. Galera nodes are avaiable directly then:
But everything related to maxscale hangs:
|
| Comments |
| Comment by markus makela [ 2018-04-26 ] |
|
This would appear to be caused by the same problem that caused |
| Comment by markus makela [ 2018-04-26 ] |
|
A fix can be found on the 2.2-markusjm branch on GitHub. The commit that fixes the bug is e5f26e2980ff727481ee1640df0146310ab48bf6. |