Many unauthenticated user thread after start slave

MySQL & MariaDB Percona Server for MySQL 8.0
1

I have a Percona mysql8.0.23 cluster, 1 master with 2 slaves on Ubuntu20, after I start io thread on slave, there are many unauthenticated user on the master, once I stop io thread, they are immediately gone, any ideas how can I troubleshoot? I set skip-name-resolve and there is no unknown ip access master, after I switch master to another machine, the problem still exists. Replication does not use ssl

    Until_Log_File: 
            Until_Log_Pos: 0
       Master_SSL_Allowed: No
       Master_SSL_CA_File: 
       Master_SSL_CA_Path: 
          Master_SSL_Cert: 
        Master_SSL_Cipher: 
           Master_SSL_Key: 
    Seconds_Behind_Master: 0

Master_SSL_Verify_Server_Cert: No
Last_IO_Errno: 0
Last_IO_Error:

| 200562 | unauthenticated user | connecting host | NULL | Connect | 7846 | login | NULL | 7845709 | 0 | 0 |
| 200754 | unauthenticated user | connecting host | NULL | Connect | 7350 | login | NULL | 7349367 | 0 | 0 |
| 201778 | unauthenticated user | connecting host | NULL | Connect | 5041 | login | NULL | 5041047 | 0 | 0 |
| 201394 | unauthenticated user | connecting host | NULL | Connect | 5952 | login | NULL | 5951720 | 0 | 0 |
| 203442 | unauthenticated user | connecting host | NULL | Connect | 1266 | login | NULL | 1265409 | 0 | 0 |
| 200946 | unauthenticated user | connecting host | NULL | Connect | 6905 | login | NULL | 6904952 | 0 | 0 |
| 202674 | unauthenticated user | connecting host | NULL | Connect | 3060 | login | NULL | 3060046 | 0 | 0 |
| 201330 | unauthenticated user | connecting host | NULL | Connect | 6120 | login | NULL | 6120046 | 0 | 0 |
| 201714 | unauthenticated user | connecting host | NULL | Connect | 5160 | login | NULL | 5160037 | 0 | 0 |
| 202226 | unauthenticated user | connecting host | NULL | Connect | 4026 | login | NULL | 4025437 | 0 | 0 |
| 202482 | unauthenticated user | connecting host | NULL | Connect | 3378 | login | NULL | 3377939 | 0 | 0 |
| 203378 | unauthenticated user | connecting host | NULL | Connect | 1380 | login | NULL | 1379881 | 0 | 0 |
| 202162 | unauthenticated user | connecting host | NULL | Connect | 4188 | login | NULL | 4187940 | 0 | 0 |
| 202802 | unauthenticated user | connecting host | NULL | Connect | 2760 | login | NULL | 2760046 | 0 | 0 |
| 203058 | unauthenticated user | connecting host | NULL | Connect | 2090 | login | NULL | 2089366 | 0 | 0 |
| 203314 | unauthenticated user | connecting host | NULL | Connect | 1500 | login | NULL | 1499366 | 0 | 0 |
| 202930 | unauthenticated user | connecting host | NULL | Connect | 2430 | login | NULL | 2429366 | 0 | 0 |
| 203634 | unauthenticated user | connecting host | NULL | Connect | 840 | login | NULL | 839367 | 0 | 0 |
| 203506 | unauthenticated user | connecting host | NULL | Connect | 1145 | login | NULL | 1144937 | 0 | 0 |
| 203698 | unauthenticated user | connecting host | NULL | Connect | 668 | login | NULL | 667939 | 0 | 0 |
| 203954 | unauthenticated user | connecting host | NULL | Connect | 18 | login | NULL | 17939 | 0 | 0 |
| 203762 | unauthenticated user | connecting host | NULL | Connect | 491 | login | NULL | 490252 | 0 | 0 |
| 203890 | unauthenticated user | connecting host | NULL | Connect | 166 | login | NULL | 166012 | 0 | 0 |
±-------±---------------------±----------------±-----±--------±-------±------±-----±----------±----------±--------------+
1680 rows in set (0.01 sec)

2

Can you connect to the source server from the same replica using a regular mysql client and the same user? If so, does the same issue happen?

3

yes,I can connect the master server from the slave by using replication user, if use mysql client connects to the master, the issue can’t happen, once start io thread , unauthenticated user become more and more on master. Replication is working fine and this is a production cluster
Slave_IO_Running: Yes
Slave_SQL_Running: Yes
I use thread-pool and application use long connection

| thread_handling | pool-of-threads |
| thread_pool_high_prio_mode | transactions |
| thread_pool_high_prio_tickets | 4294967295 |
| thread_pool_idle_timeout | 60 |
| thread_pool_max_threads | 100000 |
| thread_pool_oversubscribe | 7 |
| thread_pool_size | 64 |
| thread_pool_stall_limit | 50 |
| thread_stack | 286720 |
| thread_statistics | OFF |

4

After I disabled the thread pool, the issue doesn’t happen.

5

Why does slave connection to master result in unauthenticated user even though there is still a connection that is intact between master and slave. Why will it try to reconnect again and again.

6

Please check if you are facing the bug
https://perconadev.atlassian.net/browse/PS-9614
You can check the timer thread with command below

ps -T -p $(pidof mysqld) | grep timer_th