You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@httpd.apache.org by Jess Holle <je...@ptc.com> on 2009/04/10 19:26:10 UTC
[users@httpd] mod_proxy_balancer unbalancing?
Has anyone seen mod_proxy_balancer (in 2.2.11 with mod_proxy_ajp)
becoming unbalanced in usage?
In testing we've seen it balance fine for a minutes/hours and then
seemingly forget about one of the workers -- sending all session-less
requests to another worker (we only have 2 live workers and multiple
"dead" workers in the balancer set).
Later I see requests going to the forgotten worker again and I see
nothing (with LogLevel set to debug) to denote that Apache ever found
the worker unresponsive and thus put it in an error state.
Clearly I have more debugging ahead of me, but I just thought I'd check
to see if such an issue has already surfaced.
--
Jess Holle
P.S. Yes, I could try mod_jk instead, but mod_jk lacks one critical
capability -- the ability to throttle requests by queuing them up in
Apache rather than simply returning a 503 when the AJP workers are
max'ed out.
---------------------------------------------------------------------
The official User-To-User support forum of the Apache HTTP Server Project.
See <URL:http://httpd.apache.org/userslist.html> for more info.
To unsubscribe, e-mail: users-unsubscribe@httpd.apache.org
" from the digest: users-digest-unsubscribe@httpd.apache.org
For additional commands, e-mail: users-help@httpd.apache.org
[users@httpd] Re: mod_proxy_balancer unbalancing?
Posted by Jess Holle <je...@ptc.com>.
I'm not encouraged thus far.
I note that the busy and lbstatus counts for "dead" workers increase
without bound over time due to retries of these workers wherein busy and
lbstatus are both incremented but never decremented. This will
certainly cause bad behavior when any one of these dead workers wakes up!
This does not really explain how the 2 live workers become unbalanced,
but I do note that when the live worker that is inappropriately skipped
for a long interval is finally revisited its lbstatus is -382!
--
Jess Holle
Jess Holle wrote:
> P.S. I'm using the by-requests balancing algorithm. Perhaps I
> shouldn't be?
>
> Jess Holle wrote:
>> Has anyone seen mod_proxy_balancer (in 2.2.11 with mod_proxy_ajp)
>> becoming unbalanced in usage?
>>
>> In testing we've seen it balance fine for a minutes/hours and then
>> seemingly forget about one of the workers -- sending all session-less
>> requests to another worker (we only have 2 live workers and multiple
>> "dead" workers in the balancer set).
>>
>> Later I see requests going to the forgotten worker again and I see
>> nothing (with LogLevel set to debug) to denote that Apache ever found
>> the worker unresponsive and thus put it in an error state.
>>
>> Clearly I have more debugging ahead of me, but I just thought I'd
>> check to see if such an issue has already surfaced.
>>
>> --
>> Jess Holle
>>
>> P.S. Yes, I could try mod_jk instead, but mod_jk lacks one critical
>> capability -- the ability to throttle requests by queuing them up in
>> Apache rather than simply returning a 503 when the AJP workers are
>> max'ed out.
>
---------------------------------------------------------------------
The official User-To-User support forum of the Apache HTTP Server Project.
See <URL:http://httpd.apache.org/userslist.html> for more info.
To unsubscribe, e-mail: users-unsubscribe@httpd.apache.org
" from the digest: users-digest-unsubscribe@httpd.apache.org
For additional commands, e-mail: users-help@httpd.apache.org
[users@httpd] Re: mod_proxy_balancer unbalancing?
Posted by Jess Holle <je...@ptc.com>.
P.S. I'm using the by-requests balancing algorithm. Perhaps I shouldn't be?
Jess Holle wrote:
> Has anyone seen mod_proxy_balancer (in 2.2.11 with mod_proxy_ajp)
> becoming unbalanced in usage?
>
> In testing we've seen it balance fine for a minutes/hours and then
> seemingly forget about one of the workers -- sending all session-less
> requests to another worker (we only have 2 live workers and multiple
> "dead" workers in the balancer set).
>
> Later I see requests going to the forgotten worker again and I see
> nothing (with LogLevel set to debug) to denote that Apache ever found
> the worker unresponsive and thus put it in an error state.
>
> Clearly I have more debugging ahead of me, but I just thought I'd
> check to see if such an issue has already surfaced.
>
> --
> Jess Holle
>
> P.S. Yes, I could try mod_jk instead, but mod_jk lacks one critical
> capability -- the ability to throttle requests by queuing them up in
> Apache rather than simply returning a 503 when the AJP workers are
> max'ed out.
---------------------------------------------------------------------
The official User-To-User support forum of the Apache HTTP Server Project.
See <URL:http://httpd.apache.org/userslist.html> for more info.
To unsubscribe, e-mail: users-unsubscribe@httpd.apache.org
" from the digest: users-digest-unsubscribe@httpd.apache.org
For additional commands, e-mail: users-help@httpd.apache.org