[mod_cluster-issues] [JBoss JIRA] (MODCLUSTER-636) When setting PersistSlots to On, EAP node can not rejoin mod_cluster due to java.lang.IllegalArgumentException

[Alex Cai (JIRA)]

    [ https://issues.jboss.org/browse/MODCLUSTER-636?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13507693#comment-13507693 ] 

Alex Cai edited comment on MODCLUSTER-636 at 12/21/17 12:52 AM:
----------------------------------------------------------------

[~jfclere] We got the following feedback from customer regarding ThreadsPerChild:
--------
1. We know that ThreadsPerChild is changed frequently when using apache.
- Does it mean that we should not change the ThreadsPerChild value, if we use the PersistSlots On option.
- web server administrator said that there was a problem when only restarting the web server, and after tracking what he did, he modified the 00-mpm.conf file and confirmed that he restarted it.
- So, to reproduce, 00-mpm.conf was modified and tried to test, and reproduced more than a few times.

2. We modified the ThreadPerChild option to increase the number of threads per apache child process.
- It is set to a low value before the load test, and it is set to a high value in the load test.
- We also adjusted these options to reduce the number of child processes to find the appropriate values, and to increase the number of threads in each child process.
- If you set it to a value greater than or equal to 64, we know that it usually should be set the ThreadLimit and ThreadsPerChild values the same.
--------

1. Does this match your ananlysis?
2. Could you please share more details regarding why changing ThreadsPerChild with PersistSlots On will cause the problem? We will need to let customer understand the reason.
3. Customer also wants to know all the other restrictions besides changing the ThreadsPerChild when having PersistentSlots On. Could you please list these kind of restrictions?


Thanks a lot



was (Author: alexcred):
[~jfclere] We got the following feedback from customer regarding ThreadsPerChild:
--------
1. We know that ThreadsPerChild is changed frequently when using apache.
- Does it mean that we should not change the ThreadsPerChild value, if we use the PersistSlots On option.
- web server administrator said that there was a problem when only restarting the web server, and after tracking what he did, he modified the 00-mpm.conf file and confirmed that he restarted it.
- So, to reproduce, 00-mpm.conf was modified and tried to test, and reproduced more than a few times.

2. We modified the ThreadPerChild option to increase the number of threads per apache child process.
- It is set to a low value before the load test, and it is set to a high value in the load test.
- We also adjusted these options to reduce the number of child processes to find the appropriate values, and to increase the number of threads in each child process.
- If you set it to a value greater than or equal to 64, we know that it usually should be set the ThreadLimit and ThreadsPerChild values the same.
--------

Does this match your ananlysis?
Could you please share more details regarding why changing ThreadsPerChild with PersistSlots On will cause the problem? We will need to let customer understand the reason.

Thanks a lot


> When setting PersistSlots to On, EAP node can not rejoin mod_cluster due to java.lang.IllegalArgumentException
> --------------------------------------------------------------------------------------------------------------
>
>                 Key: MODCLUSTER-636
>                 URL: https://issues.jboss.org/browse/MODCLUSTER-636
>             Project: mod_cluster
>          Issue Type: Bug
>          Components: Native (httpd modules)
>    Affects Versions: 1.3.5.Final
>         Environment: * JBCS httpd 2.4.23 SP2
> * mod_cluster: 1.3.5
> * EAP: 7.0.8
>            Reporter: Alex Cai
>            Assignee: Jean-Frederic Clere
>
> Customer set 'PersistSlots On' and sometimes the jboss eap node could not rejoing into mod_cluster due to java.lang.IllegalArgumentException
> When they got this exception, they have to delete every cache files then restart httpd and jboss for join correctly.
> Customer tried the below actions:
> 1) apache stop -> delet manager.* files in the mod_cluster directory -> apache start 
> - this action could cause http 503 error
> or
> 2) pre-backup the all files(normal files before issue) in the mod_cluster directory. when the issue occure, apache stop -> restore manager.* files from backuped files -> apache start
> - no http 503 error
> Please see the case for more details:
> https://access.redhat.com/support/cases/01986243



--
This message was sent by Atlassian JIRA
(v7.5.0#75005)