Mantis - Resin
Viewing Issue Advanced Details
ID: Category: Severity: Reproducibility: Date Submitted: Last Update:
2862 minor always 08-25-08 09:06 08-31-09 14:41
Reporter: posenato Platform: Intel(R) Xeon(R) CPU X7350 @ 2  
Assigned To: ferg OS: Linux  
Priority: normal OS Version: RedHat AS 5.2  
Status: closed Product Version: 3.1.6  
Product Build: Resolution: fixed  
Projection: none      
ETA: none Fixed in Version: 4.0.2  
Summary: 0002862: Apache + Resin and 503 HTTP Error
Description: We have a Resin + Apache 2.2.3 system that manages some virtual web sites.

Since the load is increasing, we have updated the h/w and the Resin version.

Now it is 3.1.6.

Since the update, we have noted an increasing of the 503 httpd error. The server load is near to 0!!

We have tried to tune the timeout & keepalive parameters, but the documentation is conflicting.
The page http://www.caucho.com/resin/doc/tuning.xtp [^] has some wrong parameter names and value.

Is there any place where the relation between load-balance-idle-time and socket-timeout and keepalive-timeout is made clear?

I have read in post 0002715 that there is another parameter to set up (<load-balance-socket-timeout>), but I have no found any documentation about it and how to relate its value w.r.t. the others parameter!
Steps To Reproduce:
Additional Information:
Relationships
Attached Files:

Notes
(0003375)
Aatos   
09-03-08 02:42   
We're having the same problem on RHEL3 with Apache 2.2.8/Resin 3.1.6 loadbalanced with mod_caucho.

We get random 503 responses which do not correlate on the load state of Apache server or Resin nodes, tweaking settings seems to have no effect.
(0003523)
smallufo   
10-29-08 16:51   
I have the same situation.
In resin 3.0.x , it never occurred.
After upgrading to 3.1.6 , it happens so often.
The load is very low....

servers :

linux 2.6.9-67.0.22.ELsmp (64 bit)
CentOS release 4.7 (Final)
httpd-2.0.52-41.ent.centos4


linux 2.6.18-92.1.6.el5 (32 bit)
CentOS release 5.2 (Final)
httpd-2.2.3-11.el5_1.centos.3
(0003525)
smallufo   
10-31-08 00:52   
I think the Severity should be MAJOR
(0004024)
stone_111   
05-15-09 05:22   
Any updates?

See this problem in 3.1.8, Apache 2, SUSE, singleserver.

Is there some unknown relation between apache and resin via mod_caucho that isn't properly documented?
(0004030)
stone_111   
05-25-09 00:48   
This seems to be an issue when upgrading from older Resin versions and using a newer Apache as well. A <socket-timeout>180s</socket-timeout> under <server> seems to have remedied the problem. Seems communication between apache and resin has been changed somehow or a default value has been changed. Documentation is a bit sparse and contradicting, but we see no 503:s at all now.
(0004206)
ferg   
08-31-09 14:41   
server/26e2

The important configuration are load-balance-socket-timeout, socket-timeout, load-balance-idle-time and keepalive-timeout. The two load-balance-* configure mod_caucho. socket-timeout and keepalive-timeout configure the Resin side.

load-balance-socket-timeout needs to be large enough to handle any server-side processing delays before the first data.

socket-timeout and keepalive-timeout need to be larger than load-balance-idle-time because they need to keep the connection open for a keepalive request.

For a diagram, see

  http://caucho.com/resin/admin/load-balancing.xtp [^]