[MXS-4638] Error msg reports write query failure for read queries Created: 2023-06-13  Updated: 2023-06-14  Resolved: 2023-06-14

Status: Closed
Project: MariaDB MaxScale
Component/s: readwritesplit
Affects Version/s: 23.02.1
Fix Version/s: N/A

Type: Bug Priority: Trivial
Reporter: Clint Moyer (Inactive) Assignee: markus makela
Resolution: Incomplete Votes: 0
Labels: None


 Description   

MaxScale error message for delayed_retry_timeout does not distinguish between read/write queries.

https://github.com/mariadb-corporation/MaxScale/blob/maxscale-23.02.1/server/modules/routing/readwritesplit/rwsplit_route_stmt.cc#L954

This was observed during SkySQL testing where the following statement was executed every 2 seconds while the primary was terminated.

SELECT * FROM sampledb.employees;
 
+----+-----------+----------+-------------------------+------+
 
| id | firstname | lastname | email                   | age  |
 
+----+-----------+----------+-------------------------+------+
 
|  1 | john      | doe      | john.doe@example.com    |   30 |
 
|  2 | jane      | doe      | jane.doe@example.com    |   28 |
 
|  3 | alice     | smith    | alice.smith@example.com |   35 |
 
|  4 | bob       | johnson  | bob.johnson@example.com |   40 |
 
+----+-----------+----------+-------------------------+------+

We observed the write query error message while only testing reads. (IP address has been changed for privacy)

2023-06-13 12:36:47 warning: (41) [readwritesplit] [Read-Write-Service] Write query received from dbpgf14557113@192.0.2.0. 'delayed_retry_timeout' exceeded before a primary could be found. Closing client connection.



 Comments   
Comment by markus makela [ 2023-06-13 ]

Was that inside of a transaction or not? Are you able to provide MaxScale logs with log_info enabled?

Comment by Clint Moyer (Inactive) [ 2023-06-14 ]

I enabled info logs and tried to re-produce several times without success. Was unable to trigger the delayed_retry_timeout error again while trying a variety of ways to terminate replicas. My guess something else in SkySQL made a connection that I wasn't aware of which could have been a write. Can probably close this as a non-issue, I will keep playing with it a bit more and follow-up if I see anything.

Comment by markus makela [ 2023-06-14 ]

OK, if you find a way to reproduce it, we'll reopen this one.

Generated at Thu Feb 08 04:30:05 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.