[MXS-1976] MaxAdmin Shutting Down A Service should specify / warn that new session requests are neither accepted nor denied. Created: 2018-07-12 Updated: 2020-08-25 Resolved: 2018-11-15 |
|
| Status: | Closed |
| Project: | MariaDB MaxScale |
| Component/s: | Documentation |
| Affects Version/s: | 2.2.11 |
| Fix Version/s: | 2.3.1 |
| Type: | New Feature | Priority: | Major |
| Reporter: | Juan | Assignee: | markus makela |
| Resolution: | Fixed | Votes: | 0 |
| Labels: | None | ||
| Environment: |
CentOS |
||
| Sprint: | MXS-SPRINT-70 |
| Description |
|
When
is in progress, MaxScale continues to service current connections but does not accept new ones. This is documented in the knowledge-base here. The problem is that new connection requests are not denied , nor are they accepted. They simply hang. It would not be a problem if the default ConnectTimeout in Connector/J were not infinite/undefined, but the combination of default behaviors means that clients rather than being actively denied are left in limbo when trying to connect to a MaxScale service in the process of shutting down. One solution might be to edit the documentation to warn users that MaxScale does not actively refuse new connections when it's in the process of shutting down, instead simply not answering, letting them know that some connectors do not time out unless configured to do so. Another alternative might be to have the service actively reject new connections when it is in the process of shutting down. |
| Comments |
| Comment by markus makela [ 2018-11-13 ] |
|
Updated the maxadmin documentation for 2.3 to suggest destroying the listener to immediately reject new connections. |
| Comment by markus makela [ 2018-11-15 ] |
|
Amended the 2.3 documentation with a note about destroy listener being an option. |