Hi all,
Where i want to arrive:
1) having two storage server replicating partition with DRBD 2) exporting via GNBD from the primary server the drbd with GFS2 3) inporting the GNBD on some nodes and mount it with GFS2
Assuming no logical error are done in the last points logic this is the situation:
Server 1: LogVol09, DRDB configured as /dev/drbd0 replicated to Server 2. DRBD seems to work perfect as cat /proc/drbd Server 2: LogVol09, Secondary for /dev/drbd, cat /proc/drbd is ok
What i thought to do was to export via GNBD the /dev/drbd0 to import it on a couple of nodes. Using GFS2 as FS for concurrent usage.
I got to format the GFS2 FS on the Server 1 drbd0 device. But was unable to mount it because gfs_controld wasn't accepting connections, i discovered that cman manage gfs_controld. In any case i started it manually and discover that it try to use ccsd daemon which i have not installed.
Ok, i do not need to mount it locally so i started configuring GNBD, started manualli the server end stated the export command...again the cman/ccsd problem:
gnbd_clusterd: ERROR cannot join cman
Now, what i'm asking: am i forced to use all the cluster stuff cman, ccsd...fence...with this configuration? I would like only to use the tools i need: DRDB, GNBD and GFS2 without cluster facilities.
Thanks Raffaele
Hi,
yes, you need to go with the Cluster Stuff...
Regarding your Setup, i got the best experience with drbd + gfs + iscsi export. GNBD was not as Stable than i expected. Overall Performance was even worse, too compared with iSCSI.
Greetings
Juergen
On 03/27/2010 05:39 PM, Raffaele Camarda wrote:
Hi all,
Where i want to arrive:
- having two storage server replicating partition with DRBD
- exporting via GNBD from the primary server the drbd with GFS2
- inporting the GNBD on some nodes and mount it with GFS2
Assuming no logical error are done in the last points logic this is the situation:
Server 1: LogVol09, DRDB configured as /dev/drbd0 replicated to Server 2. DRBD seems to work perfect as cat /proc/drbd Server 2: LogVol09, Secondary for /dev/drbd, cat /proc/drbd is ok
What i thought to do was to export via GNBD the /dev/drbd0 to import it on a couple of nodes. Using GFS2 as FS for concurrent usage.
I got to format the GFS2 FS on the Server 1 drbd0 device. But was unable to mount it because gfs_controld wasn't accepting connections, i discovered that cman manage gfs_controld. In any case i started it manually and discover that it try to use ccsd daemon which i have not installed.
Ok, i do not need to mount it locally so i started configuring GNBD, started manualli the server end stated the export command...again the cman/ccsd problem:
gnbd_clusterd: ERROR cannot join cman
Now, what i'm asking: am i forced to use all the cluster stuff cman, ccsd...fence...with this configuration? I would like only to use the tools i need: DRDB, GNBD and GFS2 without cluster facilities.
Thanks Raffaele
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
Thanks Juergen for your response. I did not post till now because i've been fighting with all cluster stuff ! :D
I mean GFS2. DRBD, mysql and heartbeat work fine in an active/passive configuration.
What really does not fit to my needs is the cluster stuff which i have to use only to be able to mount the GFS2 FS.
I really do not get why i cannot be left free to choose my configuration. So to mount a GFS2 FS i had to install the cluster groups, ok i did surrended to this.
Now when i switch from one to another mysql server it doesn't start the service because of fencing! Ok i do not need fencing, but ok, i tought because it needs it i will set up a xen fence, no i can't beacause the dom0 have to be cluster members and i do not need aand want cluster stuff for dom0...
So i'm escalating problems because i cannot mount a GFS2 FS without all the cluster stuff. I will need to modifice sbin and to write script by myself to use fence_ack_manual, ok it is not suggested but i'm assuming the risk for that.
So if someone have been able to set up a GFS2 mount without Cluster stuff (cman) i will be happy to listen to!
For now i can tell great DRBD and GFS2: two great peace of software.
Cheers Raffaele
2010/3/29 Juergen Gotteswinter jg@internetx.de
Hi,
yes, you need to go with the Cluster Stuff...
Regarding your Setup, i got the best experience with drbd + gfs + iscsi export. GNBD was not as Stable than i expected. Overall Performance was even worse, too compared with iSCSI.
Greetings
Juergen
On 03/27/2010 05:39 PM, Raffaele Camarda wrote:
Hi all,
Where i want to arrive:
- having two storage server replicating partition with DRBD
- exporting via GNBD from the primary server the drbd with GFS2
- inporting the GNBD on some nodes and mount it with GFS2
Assuming no logical error are done in the last points logic this is the situation:
Server 1: LogVol09, DRDB configured as /dev/drbd0 replicated to Server 2. DRBD seems to work perfect as cat /proc/drbd Server 2: LogVol09, Secondary for /dev/drbd, cat /proc/drbd is ok
What i thought to do was to export via GNBD the /dev/drbd0 to import it
on a
couple of nodes. Using GFS2 as FS for concurrent usage.
I got to format the GFS2 FS on the Server 1 drbd0 device. But was unable
to
mount it because gfs_controld wasn't accepting connections, i discovered that cman manage gfs_controld. In any case i started it manually and discover that it try to use ccsd daemon which i have not installed.
Ok, i do not need to mount it locally so i started configuring GNBD,
started
manualli the server end stated the export command...again the cman/ccsd problem:
gnbd_clusterd: ERROR cannot join cman
Now, what i'm asking: am i forced to use all the cluster stuff cman, ccsd...fence...with this configuration? I would like only to use the
tools i
need: DRDB, GNBD and GFS2 without cluster facilities.
Thanks Raffaele
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
-
Juergen Gotteswinter -
Raffaele Camarda