MDEV-35570 parallel slave ALTER-SEQUNCE attemted to binlog out-of-order #4043
Conversation
|
|
| # | ||
| --connection slave | ||
| --source include/stop_slave.inc | ||
| --let $saved_mode= @@global.slave_parallel_mode |
There was a problem hiding this comment.
This looks wrong. The $saved_mode will be a string containing "@@glob...", not the actual value of the slave_parallel_mode which was probably intended.
Use the normal way: SET @saved_mode= @@global.slave_parallel_mode . This also avoids getting the value into the .result file when restoring, causing the test to fail if run with some non-standard --mysqld=--slave-parallel-mode option
| # allow to proceed to sync with the 1st following WFPT2SC wait condtion | ||
| lock table ti write; | ||
| # allow to proceed into commit to sync with the 2nd following WFPC wait condition | ||
| --let $saved_dbug= @@GLOBAL.debug_dbug |
There was a problem hiding this comment.
Again, use: SET @saved_dbug= @@GLOBAL.debug_dbug
…rder _(This is the 1st of two commits to reproduce the issue; the 2nd commits that follow fixes it)_ Since MDEV-31503 fixes ALTER-SEQUENCE might be able to complete its work including binlogging before a preceding in binlog-order transaction. There may not be data dependency between the two transactions, but there would be "an attempt was made to binlog GTID D-S-XYZ which would create an out-of-order sequence number" error in the gtid_strict_mode = ON. After the preceding transaction started committing, and does is very slow, ALTER-SEQUNCE was able to find a time window to complete because it temporarily releases its link with the waitee parent transaction. And while having released it it also erroneously executed binlogging part. Fixed with restoring the commit dependency on the parent before ALTER-SEQUNCE takes on binlogging.
andrelkin
force-pushed
the
bb-10.6-MDEV-35570
branch
from
0a08f55 to
f195bef
Compare
The description is copied from the commit message of the initial branch's commit
Since MDEV-31503 fixes ALTER-SEQUENCE might be able to complete its
work including binlogging before a preceding in binlog-order
transaction. There may not be data dependency between the two
transactions, but there would be
"an attempt was made to binlog GTID D-S-XYZ which would create an
out-of-order sequence number"
error in the gtid_strict_mode = ON.
After the preceding transaction started committing, and does is very
slow, ALTER-SEQUNCE was able to find a time window to complete because
it temporarily releases its link with the waitee parent transaction.
And while having released it it also erroneously executed binlogging part.
Fixed with restoring the commit dependency on the parent before
ALTER-SEQUNCE takes on binlogging.