[MDEV-7307] unauthenticated user while using pool of threads Created: 2014-12-11  Updated: 2019-09-03  Resolved: 2019-09-03

Status: Closed
Project: MariaDB Server
Component/s: OTHER
Affects Version/s: 5.5
Fix Version/s: N/A

Type: Bug Priority: Major
Reporter: Alex Assignee: Sergei Golubchik
Resolution: Not a Bug Votes: 1
Labels: None
Environment:

centos 6.x


 Description   

Hello,
Recently we've upgraded all our mysql servers to mariadb 5.5, and now started using pool of threads on a few systems. Systems are different, but all have MyISAM.

Unlike in thread-per-connection mode, while using pool of threads I see many unauthenticated connections in login state. Looks like it's 2nd phase and server awaits for final (authentication) connection after initial handshake been made. The connections come and immediately disappear. I checked with mysqlsniffer and couldn't find anything strange, everything works and looks like it's a normal behavior. Still I don't know if it's a normal thing : to see any "bad" states in processlist. So based on mysqlsniffer, all these threads numbers I see for unauth users are normally finishing. No aborted clients or any other error. Servers doing around 2-8 qps (under low load I don't see anyproblems). We use default settings for pool of threads, it's only enabled and using cpu number as num of pools.

So, wanted to ask if that's a normal behavior or not? (couldn't find anything anywhere). Of course resolving disabled...

Thanks!
Alex

| 5455148 | unauthenticated user | connecting host | NULL   | Connect  |  NULL | login    | NULL   |    0.000 |
 
| 5455163 | unauthenticated user | connecting host | NULL   | Connect  |  NULL | login    | NULL   |    0.000 |
 
| 5455172 | unauthenticated user | connecting host | NULL   | Connect  |  NULL | login    | NULL   |    0.000 |
 
| 5455187 | unauthenticated user | connecting host | NULL   | Connect  |  NULL | login    | NULL   |    0.000 |
 
| 5455196 | unauthenticated user | connecting host | NULL   | Connect  |  NULL | login    | NULL   |    0.000 |
 
| 5455211 | unauthenticated user | connecting host | NULL   | Connect  |  NULL | login    | NULL   |    0.000 |
 
| 5455220 | unauthenticated user | connecting host | NULL   | Connect  |  NULL | login    | NULL   |    0.000 |
 
| 5455235 | unauthenticated user | connecting host | NULL   | Connect  |  NULL | login    | NULL   |    0.000 |
 
| 5455244 | unauthenticated user | connecting host | NULL   | Connect  |  NULL | login    | NULL   |    0.000 |
 
| 5455259 | unauthenticated user | connecting host | NULL   | Connect  |  NULL | login    | NULL   |    0.000 |
 
| 5455268 | unauthenticated user | connecting host | NULL   | Connect  |  NULL | login    | NULL   |    0.000 |
 
| 5455283 | unauthenticated user | connecting host | NULL   | Connect  |  NULL | login    | NULL   |    0.000 |



 Comments   
Comment by Elena Stepanova [ 2014-12-12 ]

I think it's normal if they don't hang there for long (and according your description they don't). Since a new connection needs to obtain a thread from the pool, there might be some delay comparing to thread-per-connection, so they stay in this state long enough to make it to the processlist. But I'm speaking without a real knowledge of internals, so I'll assign it to serg for the second opinion.

Comment by Sergei Golubchik [ 2019-09-03 ]

I suspect it's normal for pool-of-threads and not a bug

Generated at Thu Feb 08 07:18:36 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.