[DRBD-user] Dual primary// GFS2// Cannot mount /dev/drbd0 on second DRBD node

Daniel Meszaros spam at meszi.de
Tue Aug 9 13:51:00 CEST 2011

Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.


Kaloyan Kovachev:
> Hi,
>   manual fencing is not recommended for production clusters. You should
> really have some fencing device.
>   You may try to restart xen2, then it should rejoin the cluster properly
> or at least 'fence_ack_manual -n xen2' will be safe to answer 'absolutely'
> as that is what you did - manually fenced the node :)
Ok, I did the following...

- On "xen1": fence_ack_manual -n xen2
- Rebooted "xen2"
- Started DRBD on "xen2" as secondary
- Made "xen2" primary after the resync was finished
- Restarted cman
- Mounted /dev/drbd0 successfully
- Created a test file on each node, edited it and watched the change on 
the other node.

It appears to work fine now.

But (please excuse this collection of silly questions) the command for 
manual fencing is still "open" on "xen1". How can I return to automatic 
fencing? I am afraid canceling (Ctrl+c) the command on "xen1" could do 
something nasty.

BTW, I suppose that the cluster.conf (a copy is below the email text)  
is set up wrong - which could have caused the problem. My concerns are:

- clean_start="1" ... which I configured to get the thing running on 
"xen1" when there was still no cluster configured on "xen2"
- <fencedevice name="human" agent="fence_manual"/> ... which I took over 
from the mentioned DRBD/GFS2 manual page

What do you think?


# cat /etc/cluster/cluster.conf
<?xml version="1.0"?>
<cluster name="cluster" config_version="1">
<!-- post_join_delay: number of seconds the daemon will wait before
                         fencing any victims after a node joins the domain
        post_fail_delay: number of seconds the daemon will wait before
                         fencing any victims after a domain member fails
        clean_start    : prevent any startup fencing the daemon might do.
                         It indicates that the daemon should assume all 
                         are in a clean state to start. -->
<fence_daemon clean_start="1" post_fail_delay="0" post_join_delay="3"/>
<clusternode name="xen1" votes="1" nodeid="1">
<!-- Handle fencing manually -->
<method name="human">
<device name="human" nodename="xen1"/>
<clusternode name="xen2" votes="1" nodeid="2">
<!-- Handle fencing manually -->
<method name="human">
<device name="human" nodename="xen2"/>
<!-- cman two nodes specification -->
<cman expected_votes="1" two_node="1"/>
<!-- Define manual fencing -->
<fencedevice name="human" agent="fence_manual"/>

More information about the drbd-user mailing list