Shidb / Apache Segmentation Fault

[Aleix Dorca Josa]

Scott,

Now I see what you mean. No, I'm not sharing anything between the SP. There is no shared state at all. We have an haproxy host that redirects HTTP petitions only to one Moodle host, always to the same one.

I've tried replicating the things that are done via cron job. It was true that there was a graceful restart of Apache. That is when things go crazy, but I've just done manual graceful restarts and it works fine. Maybe it has something to do with rsync using to much memory and Apache trying to do a graceful restart.

As per the principle, I meant: keeping changes on configuration files to the minimum. Sorry about not being too clear.

I will try to execute the shibd plugin in debug mode. I'll let you know.

Thanks,

Aleix.


On 11/17/14, 4:19 PM, "Aleix Dorca Josa" <adorca at uda.ad> wrote:
>
>- Parts of the system shared? Do you refer via GlusterFS or via 
>Shibboleth. In any case the system is really simple:

You said you were sharing files between the Shibboleth installations. If 
that wasn't the case, I didn't understand what you said.

>	- GlusterFS shares a folder (moodledata) with another machine via a 
>replica. It was all installed using the GlusterFS HowTo, it worked 
>flawlessly so we did not go any deeper.
>	- Shibboleth has been configured using the same principle.

I don't know what that principle is.

> We only *protect* one file (the one moodle requires) with the apache 
>plugin. On the shibboleth.xml we only changed the idp host address, and 
>the entityID of the ApplicationDefaults section.

Ok. Then you're not sharing a shibd process or trying to cluster the SP 
itself, which is a major piece of information. So there shouldn't be *any* 
shared state between the machines that's relevant to the SP, and if there 
is, you should stop that and see what happens.

>I'm dead sure the problem IS with my system. I was just asking for some 
>help. Sorry about that. Didn't meant to waste your time. :-(

Well, at the end of the day, the only way to deal with a crash is to debug 
it or get a stack trace.

-- Scott