MySQL Bugs: #86691: all threads are hanged in group replication

Bug #86691	all threads are hanged in group replication
Submitted:	14 Jun 2017 7:38	Modified:	28 Jul 2017 8:13
Reporter:	zte zte	Email Updates:
Status:	No Feedback	Impact on me:	None
Category:	MySQL Server	Severity:	S3 (Non-critical)
Version:	5.7.18	OS:	CentOS
Assigned to:	MySQL Verification Team	CPU Architecture:	Any
Tags:	group replication ，thread hanged

Description:
phenomenon
mysql 5.7.18 group replication,100 threads insert data on the master node. Then,all threads are hanged. All data is not inserted after then.

stack information shown by gdb
#0  0x000000376500b5bc in pthread_cond_wait@@GLIBC_2.3.2 () from /lib64/libpthread.so.0
#1  0x00007eff487b552e in Wait_ticket<unsigned int>::waitTicket(unsigned int const&) () from /home/rdb/lib/plugin/group_replication.so

show processlist\G
| 17711 | root        | centos177:56649 | per  | Query   |  4216 | starting                                               | commit           |
| 17712 | root        | centos177:56651 | per  | Query   |  4214 | starting                                               | commit           |

show engine innodb ststus\G
MySQL thread id 17802, OS thread handle 139636232890112, query id 3138503 centos177 10.46.178.177 root starting
commit
---TRANSACTION 4516, ACTIVE (PREPARED) 4243 sec
1 lock struct(s), heap size 1136, 0 row lock(s), undo log entries 1000
MySQL thread id 17799, OS thread handle 139635293886208, query id 3138359 centos177 10.46.178.177 root starting
commit
---TRANSACTION 4515, ACTIVE (PREPARED) 4243 sec
1 lock struct(s), heap size 1136, 0 row lock(s), undo log entries 1000
MySQL thread id 17798, OS thread handle 139634896103168, query id 3138407 centos177 10.46.178.177 root starting
commit

How to repeat:
I don't know

Suggested fix:
dead lock test，and roll back those transactions.

Hi,

Were all 100 threads locked "indefinitely" or they followed trough after timeout with aborted transaction?

Can you share your config?

best regards
Bogdan

Hi,

Can you please check that you are facing a network partition?
https://dev.mysql.com/doc/refman/5.7/en/group-replication-detecting-partitions.html

On a future version, we will provide a mechanism to deal with this automatically.
Please check BUG#84727: GR: Partitioned Node Should Get Updated Status and not accept writes.

Best regards,
Nuno Carvalho

3 Nodes are all online。

No feedback was provided for this bug for over a month, so it is
being suspended automatically. If you are able to provide the
information that was originally requested, please do so and change
the status of the bug back to "Open".