Details
-
Bug
-
Status: Closed (View Workflow)
-
Major
-
Resolution: Fixed
-
10.1.10, 10.0(EOL), 10.1(EOL)
-
None
-
Linux
Description
I tried to upgrade the server 10.0.22 -> 10.1.10.
Old version was shutdown, but new failed to start:
2016-01-08 22:25:00 7f8017479700 InnoDB: Error: Column last_update in table "mysql"."innodb_table_stats" is INT UNSIGNED NOT NULL but should be BINARY(4) NOT NULL (type mismatch).
|
2016-01-08 22:25:00 7f8017479700 InnoDB: Error: Fetch of persistent statistics requested for table "mysql"."gtid_slave_pos" but the required system tables mysql.innodb_table_stats and mysql.innodb_index_stats are not present or have unexpected structure. Using transient stats instead.
|
2016-01-08 22:25:01 140188123334592 [Note] Server socket created on IP: '::'.
|
2016-01-08 22:25:01 140188123334592 [ERROR] Failed to initialize master info using_gtid
|
2016-01-08 22:25:01 140188123334592 [ERROR] Error reading master configuration
|
2016-01-08 22:25:01 140188123334592 [ERROR] Initialized Master_info from 'master-bob@002dhk.info' failed
|
2016-01-08 22:25:01 140188123334592 [ERROR] Failed to initialize master info using_gtid
|
2016-01-08 22:25:01 140188123334592 [ERROR] Error reading master configuration
|
2016-01-08 22:25:01 140188123334592 [ERROR] Initialized Master_info from 'master-bob@002dth.info' failed
|
2016-01-08 22:25:01 140188123334592 [ERROR] Failed to initialize master info using_gtid
|
2016-01-08 22:25:01 140188123334592 [ERROR] Error reading master configuration
|
2016-01-08 22:25:01 140188123334592 [ERROR] Initialized Master_info from 'master-bob@002dsg.info' failed
|
2016-01-08 22:25:01 140188123334592 [ERROR] Failed to initialize master info using_gtid
|
2016-01-08 22:25:01 140188123334592 [ERROR] Error reading master configuration
|
2016-01-08 22:25:01 140188123334592 [ERROR] Initialized Master_info from 'master-delivery.info' failed
|
2016-01-08 22:25:01 140188123334592 [Warning] Reading of some Master_info entries failed
|
2016-01-08 22:25:01 140188123334592 [ERROR] Failed to initialize multi master structures
|
2016-01-08 22:25:01 140188123334592 [ERROR] Aborting
|
I guess I can't use mysql_upgrade when the server is not running.
Attachments
Activity
MDEV-9383-100grama.sample
MDEV-9383-ck.sample MDEV-9383-miss60.sample
I've compiled the ending of the .info files in some of our clients multi-masters.
I've used the command: tail n 6 /var/lib/mysql/master*.info > /tmp/MDEV-9383-[host].sample
I think this would give you a good samples of the "damages" the bug made 
BTW, what happens here is that when a new master.info file is written, it
does not truncate the file (for performance reasons).
So if the new file happens to be shorter, extra junk is left at the
end. This has always been so, but it has become more of a problem now that
MySQL and MariaDB are each adding new lines to the file without
coordinating.
In 10.1, a line "END_MARKER" is added at the bottom to easily avoid reading
any left-over junk, but this change was deemed too dangerous for stable 10.0
(which is fine, since 10.0 is not going to add anothing more to
master.info).
But then apparently some bugs were introduced with the previously mentioned
patch, which could cause 10.1 to incorrectly interpret some cases of extra
junk left by 10.0, causing the problems seen here. In particular, an empty
line was treated as an empty "using_gtid=" line, causing the failure.
An easy work-around, until a fixed 10.1 release becomes available, is to
just remove anything after the "using_gtid=X" line (but do not remove the
newline character at the end of that line).
BTW, the issue with leaving extra stuff at the end of master.info is fixed in 10.1 with END_MARKER.