Bug #37581 failover problem
Submitted: 23 Jun 2008 12:18 Modified: 26 Jul 2008 18:38
Reporter: Benoit MOISAN Email Updates:
Status: No Feedback Impact on me:
None 
Category:MySQL Proxy: Core Severity:S3 (Non-critical)
Version:0.6.0 OS:Linux (rh4 64 bits)
Assigned to: CPU Architecture:Any
Tags: failover, Red Hat

[23 Jun 2008 12:18] Benoit MOISAN
Description:
Hi,

I've installed mysql on my architecture. I'll try to test the load-balancing and failover.I've an applicative server wich calls 2 mysql slave servers via mysql-proxy.

The load balancing seems to work.
When I turn off one of my 2 slaves, all the traffic is redirected to the other slave.

The problem I have is that when i re-turn on the slave which was down, the trafic stays on the second slave and the load balancing don't work any more.

How to repeat:
install mysql-proxy 0.6.0. and launch :

./mysql-proxy --proxy-address=192.168.75.7:4040 --proxy-backend-addresses=192.168.75.3:3306 --proxy-backend-addresses=192.168.75.6:3306

cut one slave. And restart it. The traffic is entirely on the second slave.
[23 Jun 2008 20:40] Sveta Smirnova
Thank you for the report.

I can not repeat described behavior. Please provide mysql-prosy output. In my case it looks like below:

$../../host/mysql-proxy-0.6.0-linux-rhas3-x86/sbin/mysql-proxy --proxy-backend-addresses=host:33051 --proxy-backend-addresses=host:33052
network-mysqld.c.302: connect(host:33052) failed: Connection refused
network-mysqld-proxy.c.3713: connecting to backend (host:33052) failed, marking it as down for ...
network-mysqld-proxy.c.3610: backend host:33052 was down for more than 10 sec, waking it up
[26 Jun 2008 15:08] Benoit MOISAN
We've both the same trace. If you launch mysql-administrator, you'll see the default. When you cut one slave, the other takes all the traffic. When you restart the first slave, all the traffic stays on the second slave.
[26 Jun 2008 15:21] Benoit MOISAN
I've waited a moment. And it seems that the first slave takes traffic once again. 
But less than the second. You have 4 fewer traffic than on the second.
[26 Jun 2008 18:38] Sveta Smirnova
Thank you for the feedback.

> I've waited a moment. And it seems that the first slave takes traffic once again. 

Does this mean you are not experiencing the problem and we can close the report as "Can't repeat"?
[26 Jul 2008 23:00] Bugs System
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".