You are right Bill
For newbees on Gluster it is very important to notify it.
This behaviour is very different than others clusters solutions and need
to be noticed in the documentation.
In my case this mechanism needs to change all softwares paths installed
on the computer according this prerequisite.
For example is you want to set an Open LDAP server GLuster cluster you
need to switch "/var/lib/ldap" path to the mount point (eg
"/var/lib/ldap-clustered") in the slapd.conf in all clustered server.
Currently this method is in debug mode for me because it seems there are
some files user "ownership" problem for the mount point.
The mount point change the stored files ownership to "bind" user instead
the real user.
For my Open LDAP example the /var/lib/ldap-clustered owner is changed to
"bind" instead "ldap" user so the LDAP crash definitely...
-------- Message initial --------
De: Bill Bruns <billbruns at yahoo.com>
?: David Touzeau <david at touzeau.eu>, Larry Bates
<larry.bates at vitalesafe.com>
Cc: Raghavendra G <raghavendra at gluster.com>, gluster-users
<gluster-users at gluster.org>
Sujet: Re: Faster AFR recovery; and try to understand
how it works...
Date: Sun, 27 Dec 2009 09:48:44 -0800 (PST)
David and Larry,
Of course, Raghavendra has once again given the correct answer!
However, I hope it will be acceptable if I offer a slight rewording.
Please consider this wording:
Create the files through one of the CLIENTs, through the mounted glusterfs (that is, through the /mnt/replicate).
Do not create the files directly on the SERVERs (in other words, do not create it through /home/replicate)
In David's case, the SERVERs are the machines with the "/home/replicate" directories, and the CLIENTs are the machines with the "/mnt/replicate" directories.
HTH
Bill Bruns
