Hi,
On Thu, Aug 13, 2009 at 12:32, Curt Mills<hacker@xxxxxxxxx> wrote:
> You mean like every time new Apache2 updates come along they
> _shouldn't_ break my failover cluster???
If you are using a cluster, and configuration files in a shared
directory, I believe you should configure your daemon not to use the
config files in the system directories such as /etc/httpd and to get
them from another path.
In the case of Apache, I think you could fix your issue by removing
all the symbolic links in /etc/httpd and leaving it as installed by
the RPM, and then editing /etc/sysconfig/httpd and adding something
like OPTIONS="-d /replicated/etc/httpd" or OPTIONS="-f
/replicated/etc/httpd/conf/httpd.conf" or even both.
That way, RPM upgrades should never touch the config files and
symbolic links under /replicated/etc/httpd, and even if they change
things under /etc/httpd you wouldn't care about it since you are not
using those...
The only think is that a naive sysadmin could try to edit the config
files in /etc/httpd but would not see any effects from it, since
Apache is not using those at all... But in any case, if you have a
cluster you have special procedures to restart and reload Apache
anyway, so any sysadmin should know what he's doing before messing
with something in your environment... I suggest you work around that
by creating some README.txt files in those directories explaining that
the "live" configuration files are in /replicated/etc/httpd instead.
This is certainly not a complete procedure on how to configure things
so that upgrades don't break your cluster, but I believe the ideas
outlined above could lead you there if you set up a test environment
and experiment a little bit with it.
HTH,
Filipe
_______________________________________________
CentOS mailing list
CentOS@xxxxxxxxxx
http://lists.centos.org/mailman/listinfo/centos
