Ocfs2 users - Jun 2009 - o2net_check_handshake ... advertised net

Hi Sergio,
thanks for an answer. Can I ask you for more explanation, please?
Does it mean that if I want to mount ocfs2/drbd partitions concurrently I must 
use OCFS2 drivers compiled with the same version kernel on all nodes?
And on the other hand - if I'd umount this partitions on non-upgraded node 
would I be able to mount them on upgraded one? Or need I remove drivers as 
well? I can't stop virtual guests and play with running node to try it but I
can plan and announce short break for some night if it would be the way.
Thanks in advance
Tomas Lavicky
> Date: Mon, 1 Jun 2009 08:36:41 -0300
> From: S?rgio Surkamp <sergio at gruposinternet.com.br>
> Subject: Re: [Ocfs2-users] o2net_check_handshake ... advertised net
> 	protocol version 11 but 8 is required
> To: ocfs2-users at oss.oracle.com
>
> Hi Mr. Tomas,
>
> You need the same version of OCFS2 drivers in both nodes to mount the
> volume. The OCFS2 versions are on-disk compatible but the
> network protocol can change. It seems your problem.
>
> Regards,
> S?rgio
>
> Em Mon, 1 Jun 2009 11:06:33 +0200
>
> Tomas Lavicky <tomas.lavicky at racom.eu> escreveu:
> > Hi,
> > we have two node HP Proliant DL140 / Ubuntu 8.04 LTS / Heartbeat
> > 2.1.3-2 cluster with bunch of KVM virtual guests. Images for virtual
> > servers are stored in two DRBD 8.0.11 / OCFS2 1.3.9 dual-primary mode
> > resources. I've run this cluster for one year yet and I need to
> > upgrade it now. I can't stop cluster and upgrade both nodes at
once.
> > So I moved all virtual servers to one node, stopped Heartbeat, OCFS2
> > and DRBD and upgrade another node to Ubuntu 8.10. All seems to be
> > well after reboot except I'm not able mount OCFS2/DRBD partitions
on
> > upgraded node:
> >
> > ~# mount -t ocfs2 /dev/drbd1 /drbd1/
> > mount.ocfs2: Transport endpoint is not connected while
> > mounting /dev/drbd1 on /drbd1/. Check 'dmesg' for more
information on
> > this error.
> >
> > Tail of dmesg on upgraded node:
> >
> > [329491.075605] (4849,0):o2net_check_handshake:1227 node
> > rrhb-gandalfb (num 0) at 10.123.45.18:7777 advertised net protocol
> > version 8 but 11 is required, disconnecting
> > [329495.106007] (7751,0):dlm_request_join:1033 ERROR: status = -107
> > [329495.106055] (7751,0):dlm_try_to_join_domain:1207 ERROR: status
> > -107 [329495.106086] (7751,0):dlm_join_domain:1485 ERROR: status >
> -107 [329495.106164] (7751,0):dlm_register_domain:1732 ERROR: status
> > = -107 [329495.106198] (7751,0):o2cb_cluster_connect:302 ERROR:
> > status = -107 [329495.106229] (7751,0):ocfs2_dlm_init:2756 ERROR:
> > status = -107 [329495.106268] (7751,0):ocfs2_mount_volume:1208 ERROR:
> > status = -107 [329495.106323] ocfs2: Unmounting device (147,1) on
> > (node 0)
> >
> > And on non-upgraded node:
> >
> > [1626792.088584] (4888,0):o2net_check_handshake:1144 node
> > rrhb-gandalfs (num 1) at 10.123.45.19:7777 advertised net protocol
> > version 11 but 8 is required, disconnecting
> >
> > Along to this thread
> > http://oss.oracle.com/pipermail/ocfs2-users/2008-April/002769.html it
> > could be kernel incompatibility problem. Is it chance to solve it
> > anyhow with different kernels? I need to keep one node running
> > continually so I'm afraid to upgrade it while the other one
won't be
> > OK. Thanks for your advice.
> >
> > Tomas

Em Mon, 1 Jun 2009 15:39:15 +0200
Tomas Lavicky <tomas.lavicky at racom.eu> escreveu:
> Hi Sergio,
> thanks for an answer. Can I ask you for more explanation, please?
> Does it mean that if I want to mount ocfs2/drbd partitions
> concurrently I must use OCFS2 drivers compiled with the same version
> kernel on all nodes?
Dunno about drbd, but for OCFS2 that's right.
> And on the other hand - if I'd umount this
> partitions on non-upgraded node would I be able to mount them on
> upgraded one?
That's right too. You can mount the upgraded node version or the old
version node, but can't mount both in cluster mode (concurrently) due
to network protocol incompatibility, as it can change between releases.

The on-disk data structures don't change between releases, so its safe
to umount the volume with the old driver, mount with the newer driver,
test anything you want and rollback to the old, by umounting the new
and mounting the old, if you find some trouble.

At my job we have done it to change from version 1.2 to 1.4 and it was a
flawless and sweet upgrade.

Regards,
-- 
  .:''''':.
.:'        `     S?rgio Surkamp | Gerente de Rede
::    ........   sergio at gruposinternet.com.br
`:.        .:'
  `:,   ,.:'     *Grupos Internet S.A.*
    `: :'        R. Lauro Linhares, 2123 Torre B - Sala 201
     : :         Trindade - Florian?polis - SC
     :.'
     ::          +55 48 3234-4109
     :
     '           http://www.gruposinternet.com.br