[MDEV-6060] MySQL BUG#13504281 - FOLLOW UP 'LIVE UPGRADE FAILS BETWEEN 5.5 -> 5.6.4 (BUG#13427444) Created: 2014-04-10  Updated: 2022-11-02  Resolved: 2022-11-02

Status: Closed
Project: MariaDB Server
Component/s: Replication
Affects Version/s: 10.0.10
Fix Version/s: N/A

Type: Bug Priority: Minor
Reporter: Sergey Vojtovich Assignee: Kristian Nielsen
Resolution: Not a Bug Votes: 0
Labels: None

Issue Links:
PartOf
is part of MDEV-4784 merge test cases from 5.6 Stalled

 Description   

Test case for this bug (disabled_replication.test) fails in 10.0. Please check if we're affected by this problem.

revno: 3690.1.74
 
committer: Alfranio Correia <alfranio.correia@oracle.com>
 
branch nick: mysql-trunk
 
timestamp: Wed 2012-03-14 09:52:49 +0000
 
message:
 
  BUG#13504281 - FOLLOW UP 'LIVE UPGRADE FAILS BETWEEN 5.5 -> 5.6.4 (BUG#13427444)
 
 
 
  Replication objects are always created so repositories are checked in the
 
  initialization routine even when --server-id is equal to "0" (i.e. default
 
  value), although a valid --server-id (i.e. not equal to the default value)
 
  is a requirement to use either a master or slave.
 
 
 
  This may cause confusion in live upgrades when an user tries to migrate
 
  from 5.5 to 5.6 because the operation stops with an error and asking users
 
  to provide --skip-slave-start in order to avoid the problem.
 
 
 
  To avoid this scenario, we only create the replication objects when a valid
 
  --server-id is provided.



 Comments   
Comment by Sergei Golubchik [ 2015-02-21 ]

This test verifies that with --server-id=0 replication commands from the list below fail with an error:

RESET SLAVE
 
STOP SLAVE
 
START SLAVE
 
CHANGE MASTER ...
 
SHOW MASTER LOGS
 
RESET MASTER

Comment by Elena Stepanova [ 2015-02-23 ]

We don't seem to have the initial problem with upgrade 5.5=>10.x (maybe because MariaDB doesn't create those kinds of replication objects, or does it differently?)
MariaDB also doesn't throw errors on the commands above, except for CHANGE MASTER. We definitely shouldn't make them fail in 10.0.
So, probably the test case should be disabled.

Comment by Sergey Vojtovich [ 2015-02-24 ]

Kristian, could you check this too? We don't have this test case, so there is nothing to disable. Otherwise I tend to agree with Elena.

Generated at Thu Feb 08 07:09:04 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.