Are you configuring ocfs2 with the user-space driven heartbeat set to
yes and driven by heartbeat/crm ?

I've crashed a couple of ocfs2 clusters, they were inactive thou...,
testing whether NodeB would survive NodeA's crash, during other
situations NodeB wouldn't crash but I/O would be frozen for a couple
of minutes while ocfs2 figured out what happened, that all went away
as soon as I put ocfs2 under heartbeat control.

regards,
esv.




On Mon, Jan 25, 2010 at 7:33 PM, Hunny Bunny <[email protected]> wrote:
> Hello Enrique,
> Thanks for your prompt reply.
> Trying to answer you question.
>
> In my configuration both nodes have hardware (Promise EX8350) RAID5 of eight 
> SATA hard drives available to the each system as /dev/sda
>
> On the both nodes /dev/sda has four identical partitions:
>
> /dev/sda1 as swap
> /dev/sda2 as /
> /dev/sda3/ mounted as /conf
> /dev/sda4 on which /dev/drbd0 is created
>
> So, practically each of two hosts or nodes have physically attached SATA data 
> storage.
>
> DRBD block device /dev/drbd0 on top of /dev/sda4 is being replicated in dual 
> master mode between node1 and node2.
>
> OCFS2 is created atop of /dev/drbd0 and thus shared using DLM between these 
> two nodes and mounted as /data on the each node.
>
> Regards,
>
> Alex
>
>
> --- On Mon, 1/25/10, Enrique Sanchez <[email protected]> 
> wrote:
>
> From: Enrique Sanchez <[email protected]>
> Subject: Re: [Linux-HA] Strange problem with OCFS2 on top of DRBD during 
> hardware failure.
> To: "General Linux-HA mailing list" <[email protected]>
> Date: Monday, January 25, 2010, 11:51 PM
>
> Alex,
>
> Can you describe what type of devices and how are they connected to
> the hosts? are they shared devices ?
>
> regards,
> esv.
>
> On Mon, Jan 25, 2010 at 2:53 PM, Hunny Bunny <[email protected]> wrote:
>> Hello folkz,
>> I have two nodes with DRBD configured as dual primary for node1 and node2..
>> /dev/drbd0 block formatted as OCFS2 and mounted as /data on the both nodes.
>> I had a raid failure on the node1 and immediately ocfs2 /data mount on the 
>> node2 became unavailable and shortly thereafter node2 went into reboot.
>> I wonder if this an is expected behavior of OCFS2 or I missed something in 
>> /etc/drbd.conf or /etc/ocfs2/cluster.conf
>>
>> Please take a look at my config files below.
>> Thank you in advance,
>> Alex
>>
>> /etc/ocfs2/cluster.conf
>> <------------------------- snipped ---------------------------->
>>
>> cluster:
>>     node_count  = 2
>>     name        = nas
>>
>> node:
>>     ip_port     = 7777
>>     ip_address  = 192.168.0.1
>>     number      = 1
>>     name        = node1
>>     cluster     = nas
>>
>> node:
>>     ip_port     = 7777
>>     ip_address  = 192.168.0.2
>>     number      = 2
>>     name        = node2
>>     cluster     = nas
>>
>> <------------------------- snipped ---------------------------->
>>
>>
>> /etc/drbd.conf
>> <------------------------- snipped ---------------------------->
>>
>> global {
>>     usage-count no;
>> }
>>
>> common {
>>     protocol C;
>>
>>     syncer { rate 100M; }
>> }
>>
>> resource "cluster" {
>>
>>     device      /dev/drbd0;
>>     disk        /dev/sda4;
>>     meta-disk   internal
>>
>> handlers {
>>
>>         pri-on-incon-degr 
>> "/usr/local/cluster/lib/drbd/notify-pri-on-incon-degr.sh; 
>> /usr/local/cluster/lib/drbd
>>
>>         pri-lost-after-sb 
>> "/usr/local/cluster/lib/drbd/notify-pri-lost-after-sb.sh; 
>> /usr/local/cluster/lib/drbd
>>         local-io-error "/usr/local/cluster/lib/drbd/notify-io-error.sh; 
>> /usr/local/cluster/lib/drbd/notify-emer
>>
>>         fence-peer "/usr/local/cluster/lib/drbd/crm-fence-peer.sh";
>>    }
>>
>>    startup {
>>
>>         wfc-timeout  0;
>>
>>         degr-wfc-timeout 120;.
>>
>>         outdated-wfc-timeout 2;
>>    .
>>         become-primary-on both;
>>     }
>>
>>     disk {
>>
>>         on-io-error   detach;
>>
>>    }
>>
>>   net {
>>
>>         allow-two-primaries;
>>
>>         after-sb-0pri discard-zero-changes;
>>
>>         after-sb-1pri discard-secondary;
>>
>>         after-sb-2pri disconnect;
>>
>>         rr-conflict disconnect;
>>
>>     }
>>
>>  syncer {
>>
>>         rate 100M;
>>
>>         al-extents 257;
>>
>>     }
>>
>>     on node1 {
>>         address     192.168.0.1:7788;
>>     }
>>
>>     on node2 {
>>         address     192.168.0.2:7788;
>>     }
>>
>> }
>>
>> <------------------------- snipped ---------------------------->
>>
>>
>>
>>
>>
>> _______________________________________________
>> Linux-HA mailing list
>> [email protected]
>> http://lists.linux-ha.org/mailman/listinfo/linux-ha
>> See also: http://linux-ha.org/ReportingProblems
>>
>
>
>
> --
> Enrique Sanchez Vela
> ------------------------------------------
> _______________________________________________
> Linux-HA mailing list
> [email protected]
> http://lists.linux-ha.org/mailman/listinfo/linux-ha
> See also: http://linux-ha.org/ReportingProblems
>
>
>
>
> _______________________________________________
> Linux-HA mailing list
> [email protected]
> http://lists.linux-ha.org/mailman/listinfo/linux-ha
> See also: http://linux-ha.org/ReportingProblems
>



-- 
Enrique Sanchez Vela
------------------------------------------
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

Reply via email to