Patrick Donker wrote:
Hi everybody,
First of all, I am new to this list and ocfs2, so forgive my ignorance.
Anyhow, what I'm doing is this:
I'm experimenting on a 2 node debian etch shared fs and have
installed ocfs2-tools 1.2.1.
The debs run on a vmware esx 3.0.0 server and are clones of a default
template.
This is my cluster.conf:
cluster:
node_count = 2
name = san node:
ip_port = 7777
ip_address = 192.168.100.2
number = 0
name = mail cluster = san node:
ip_port = 7777
ip_address = 192.168.100.5
number = 1
name = san
cluster = san
If I start and mount the fs on one of the nodes, everything goes
fine. However, as soon as I mount the fs on the other node I get a
kernel panic with this message:
Dec 17 13:06:01 san kernel: (2797,0):o2hb_do_disk_heartbeat:854
ERROR: Device "sdb": another node is heartbeating in our slot!
mounted.ocfs2 -d on both nodes tell me this:
/dev/sdb ocfs2 6616a964-f474-4c5e-94b9-3a20343a7178
fsck.ocfs2 -n /dev/sdb
Checking OCFS2 filesystem in /dev/sdb:
label: <NONE>
uuid: 66 16 a9 64 f4 74 4c 5e 94 b9 3a 20 34 3a 71 78
number of blocks: 26214400
bytes per block: 4096
number of clusters: 3276800
bytes per cluster: 32768
max slots: 16
Somehow both nodes use the same slot to heartbeat in. Not sure what
causes this or how to change this. Please help me debug this problem
because I'm stuck.
Thanks
Patrick
_______________________________________________
Ocfs2-users mailing list
[email protected]
http://oss.oracle.com/mailman/listinfo/ocfs2-users
Sunil Mushran wrote:
As per the config, your node names are 'san' and 'mail'.
Are the names the same as the hostname?
Do on both nodes:
# for i in /config/cluster/san/node/*/local ; do LOCAL=`cat $i`; if [
$LOCAL -eq 1 ] ; then echo $i; fi; done;
You should see /config/cluster/san/node/mail/local on mail and
/config/cluster/san/node/san/local on san.
For more, refer to the user's guide, faq sand the mount/umount support
guide in the doc section on http://oss.oracle.com/projects/ocfs2.
Thanks for the suggestion, but if I enter:
for i in /config/cluster/san/node//*/local ; do LOCAL=`cat $i`; if [
$LOCAL -eq 1 ] ; then echo $i; fi; done; /
as you suggested, I get
/cat: /config/cluster/san/node/*/local: No such file or directory
-bash: [: -eq: unary operator expected
/
So I guess there is either a typo in your query, or there is an issue
with my set up. I dont have enough linux knowledge (yet ;) to decide
which one it is...
The hostnames equal the node names.
Last night I added another node, just to see what happens, and to my
surprise all goes well. Now, these are my thoughts, please bear with me:
san is the vm where an iscsi target is running, it consists of a deb
installation, with 2 additional virtual hdd's (sdb & sdc), which I
export. On sdb and sdc I have created an ocfs2 fs.
On the other nodes, amongst which 'mail' is one of them, I connect to
the target using an iscsi initiator, which works fine. As soon as I
mount the new iscsi drive, and I monitor activity on 'san' using watch
-d -n 1 "echo \"hb\" | debugfs.ocfs2 -n /dev/sdb" I see a heartbeat
originating from node 0. If I do the same from another node which I've
added, 'deb01', I see another heartbeat appearing from that node.
Everything works fine so far.
Now, as soon as I mount /dev/sdb on 'san' itself, I get the 'another
node is heartbeating in our slot!' message, and the system fences all
nodes, which results in kernel panics. Apparently 'san' is trying to
heartbeat on slot 0, which already is occupied by 'mail'. Looking at the
cluster.conf, 'san' should select slot 1. How come it is trying to use 0
then?
Am I correct in assuming that I cannot mount the ocfs2 fs on the system
which is running the cluster???
-Patrick
_______________________________________________
Ocfs2-users mailing list
[email protected]
http://oss.oracle.com/mailman/listinfo/ocfs2-users
