[Ocfs2-users] OCFS2 Volume not mounted during boot

[Joseph Qi]

Hi Bernd,
The error message shows that connection between node 1 and 2 cannot be
set up. So you should make sure network is up before mounting and can
be reached through port 7777.

Thanks,
Joseph

On 2016/10/25 3:36, Lentes, Bernd wrote:
> Hi,
> 
> i have two nodes (SLES 11 SP4 64bit), which are connected via FC to a SAN. On the SAN i created an OCFS2 volume.
> One host (let's call him 20) mounts the OCFS2 volume while booting automatically. The other (let's call him 10) doesn't.
> Here is my /etc/ocfs2/cluster.conf:
> 
> cluster:
>         node_count = 2
>         name = idg
> 
> node:
>         ip_port = 7777
>         ip_address = 192.168.100.10
>         number = 1
>         name = sunhb65277
>         cluster = idg
> 
> node:
>         ip_port = 7777
>         ip_address = 192.168.100.20
>         number = 2
>         name = sunhb58820
>         cluster = idg
> 
> 
> 192.168.100.10 is host 10, 192.168.100.20 is host 20. File is identical on both nodes.
> 
> /etc/fstab:
> /dev/disk/by-id/dm-uuid-mpath-3600c0ff00012824b04af7a5201000000         /images         ocfs2           _netdev,defaults        0 0
> 
> 
> This is the error message on 10:
> 
> Oct 24 19:16:27 sunhb65277 kernel: [   46.302046] OCFS2 1.5.0
> Oct 24 19:17:23 sunhb65277 kernel: [  102.296137] (kworker/u:0,5,3):o2net_connect_expired:1724 ERROR: no connection established with node 2 after 60.0 seconds, giving up and returning errors.
> Oct 24 19:17:23 sunhb65277 kernel: [  102.296182] (mount.ocfs2,6555,0):dlm_request_join:1472 ERROR: Error -107 when sending message 510 (key 0x666c6172) to node 2
> Oct 24 19:17:23 sunhb65277 kernel: [  102.296188] (mount.ocfs2,6555,0):dlm_try_to_join_domain:1648 ERROR: status = -107
> Oct 24 19:17:23 sunhb65277 kernel: [  102.296193] (mount.ocfs2,6555,0):dlm_join_domain:1948 ERROR: status = -107
> Oct 24 19:17:23 sunhb65277 kernel: [  102.296311] (mount.ocfs2,6555,0):dlm_register_domain:2214 ERROR: status = -107
> Oct 24 19:17:23 sunhb65277 kernel: [  102.296330] (mount.ocfs2,6555,0):o2cb_cluster_connect:313 ERROR: status = -107
> Oct 24 19:17:23 sunhb65277 kernel: [  102.296334] (mount.ocfs2,6555,0):ocfs2_dlm_init:2995 ERROR: status = -107
> Oct 24 19:17:23 sunhb65277 kernel: [  102.296350] (mount.ocfs2,6555,0):ocfs2_mount_volume:1881 ERROR: status = -107
> Oct 24 19:17:23 sunhb65277 kernel: [  102.296387] ocfs2: Unmounting device (252,5) on (node 0)
> Oct 24 19:17:23 sunhb65277 kernel: [  102.296395] (mount.ocfs2,6555,0):ocfs2_fill_super:1236 ERROR: status = -107
> 
> The error is logical. In SLES, the firewall init script is the last one executed. I don't know why, but it seems to be normal for SuSE.
> So, port 7777 is not opened when host 20 tries to connect host 10. And when the port is opened, host 20 has already stopped connecting.
> The host stuck in the init script from ocfs2 until "Network idle timeout: 60000" has run out.
> The other way (host 20 booting independent if host 10 is online or not), the ocfs2 init script starts the mount, waits some seconds
> and the host continues to boot (and the ocfs2 volume is mounted).
> 
> What i find out already is that the node with the higher number (number 2, host 20) tries to connect the node with the lower number (number 1,host 10)
> (https://oss.oracle.com/pipermail/ocfs2-users/2009-June/003626.html).
> Although i would expect that always the booting host tries to connect the other one(s).
> Host 20 also mounts automatically when host 10 is offline.
> 
> Questions:
> 
> Why does host 20 mount automatically and host 10 does not ?
> From where does host 20 know that host 10 is trying to mount the ocfs2 volume,
> because in exactly that moment host 20 tries to connect host 10 on port 7777 ?
> There is no packet from host 10 visible before !?!
> 
> Of course i could fumble on the init scripts or change the order of them, but i would prefer having a running solution out of the box.
> And i like to understand it.
> 
> 
> Bernd
>