Details
-
Type:
Bug
-
Status: Closed
-
Priority:
Major
-
Resolution: Duplicate
-
Affects Version/s: 10.0.10
-
Fix Version/s: 10.0.11
-
Component/s: None
-
Labels:None
-
Environment:Linux,RHEL6
Description
Setup: dual master (probably not relevant), N slaves, everything MariaDB 10.0.10
All servers essentially the same config (copies from same source with only server specifics changed)
Binlog format MIXED
InnoDB engine
Parallel replication enabled
All charset settings utf8
Table/column utf8.
utf8 string inserted on active master, reads back ok.
On other master (acting as slave), incorrect (charset mangling)
On some slaves, OK
On other slaves, incorrect
Which slaves get a string incorrect changes per row, so it's not always the same ones.
Interesting isn't it? ![]()
Verified by doing SELECT SHA1() on the affect column, to rule out any client mangling.
SELECT string, SHA1(string) FROM t\G
Correct:
string: Atlético Madrid
SHA1(string): 6dec52f694515702664a20b7660574139267a46a
Incorrect:
string: Atlético Madrid
SHA1(string): db4bb2aa6aae25ce19e0dda3626d37fa356c673e
Gliffy Diagrams
Attachments
Issue Links
- duplicates
-
MDEV-6156 Parallel replication incorrectly caches charset between worker threads
-
- Closed
-
Activity
- All
- Comments
- Work Log
- History
- Activity
- Transitions
I realise that because of the setup this one might be tricky to reproduce.
At this point I'm particularly interested in possible ideas on how this issue could occur, as that might also help me devise a setup to reproduce the issue in an isolated environment.