Hi,
I want to share a partition of an external disk between 3 machines. It
is located on a Sun StoreTek 3510 and bound (via multipathing) to all 3
machines.
One machine (say mach1) is the node I used to create the filesystem and
setup /etc/ocfs2/cluster.conf. I copied the file to the other 2
machines. All use the same /etc/multipath.conf[1], too. All start
/etc/init.d/ocfs2 at boot, all are running with the same OS and patches
(SLES 9 SP 4) and the same version of ocfs2 tools
(ocfs2-tools-1.2.1-4.2, ocfs2console-1.2.1-4.2). The volume is mounted
to /zdisk/ocfs2 on all machines (theoretically).
When I mount the volume on mach1, it is successful. On the other
machines (mach2 and mach3), I get: "mount.ocfs2: Transport endpoint is
not connected while mounting /dev/disk/by-name/commondisk-part1 on
/zdisk/ocfs2". On mach1, mount shows "/dev/dm-4 on /zdisk/ocfs2 type
ocfs2 (rw,_netdev,heartbeat=local)".
[1] /etc/multipath.conf translates the WWID into
/dev/disk/by-name/commondisk-part1
Following my /etc/ocfs2/cluster.conf:
node:
ip_port = 7777
ip_address = 141.65.128.114
number = 0
name = mach1
cluster = ocfs2
node:
ip_port = 7777
ip_address = 141.65.128.113
number = 1
name = mach2
cluster = ocfs2
node:
ip_port = 7777
ip_address = 141.65.128.112
number = 2
name = mach3
cluster = ocfs2
cluster:
node_count = 3
name = ocfs2
What do I do wrong? How may I succes in mounting the volume on all 3
machines?
Regards,
Werner
--
Werner Flamme, Abt. WKDV
Helmholtz-Zentrum f?r Umweltforschung GmbH - UFZ
Permoserstr. 15 - 04318 Leipzig
Tel.: (0341) 235-1921 - Fax (0341) 235-451921
http://www.ufz.de - eMail: werner.flamme at ufz.de
Check dmesg. In short, the tcp connect is failing. In ocfs2, the higher node number initiates the connection between two nodes. Do: ping -I ethX ipaddress to see if the nodes can ping each other. Other things to look for would be firewalls (internal/external) interfering with traffic. See if iptables is running. If so, see if it works after shutting it down. Werner Flamme wrote:> Hi, > > I want to share a partition of an external disk between 3 machines. It > is located on a Sun StoreTek 3510 and bound (via multipathing) to all 3 > machines. > > One machine (say mach1) is the node I used to create the filesystem and > setup /etc/ocfs2/cluster.conf. I copied the file to the other 2 > machines. All use the same /etc/multipath.conf[1], too. All start > /etc/init.d/ocfs2 at boot, all are running with the same OS and patches > (SLES 9 SP 4) and the same version of ocfs2 tools > (ocfs2-tools-1.2.1-4.2, ocfs2console-1.2.1-4.2). The volume is mounted > to /zdisk/ocfs2 on all machines (theoretically). > > When I mount the volume on mach1, it is successful. On the other > machines (mach2 and mach3), I get: "mount.ocfs2: Transport endpoint is > not connected while mounting /dev/disk/by-name/commondisk-part1 on > /zdisk/ocfs2". On mach1, mount shows "/dev/dm-4 on /zdisk/ocfs2 type > ocfs2 (rw,_netdev,heartbeat=local)". > > [1] /etc/multipath.conf translates the WWID into > /dev/disk/by-name/commondisk-part1 > > Following my /etc/ocfs2/cluster.conf: > node: > ip_port = 7777 > ip_address = 141.65.128.114 > number = 0 > name = mach1 > cluster = ocfs2 > > node: > ip_port = 7777 > ip_address = 141.65.128.113 > number = 1 > name = mach2 > cluster = ocfs2 > > node: > ip_port = 7777 > ip_address = 141.65.128.112 > number = 2 > name = mach3 > cluster = ocfs2 > > cluster: > node_count = 3 > name = ocfs2 > > What do I do wrong? How may I succes in mounting the volume on all 3 > machines? > > Regards, > Werner > >
Sunil Mushran [23.04.2008 18:01]:> Check dmesg. > > In short, the tcp connect is failing. In ocfs2, the higher node > number initiates the connection between two nodes.So, I have to mount my cluster on mach3 before mounting on mach2 and mach1? My ... - now I did so and the volume ist mounted on all 3 machines. I thought mach1 was s.th. like a "leader of the pack" ;-) because it was the first node in the cluster... The problem is solved now, I think :-)> Do: ping -I ethX ipaddress to see if the nodes can ping each other.They do. The SAP systems on the machines are accessed via those addresses as well, the boxes communicate via those addresses some SAP data, and so on.> > Other things to look for would be firewalls (internal/external) > interfering with traffic. See if iptables is running. If so, see if > it works after shutting it down.No, there is no iptables rule active. No other firewall ist up. At first, I chose other network addresses, but now I chose the network addresses that are associated with the hostnames in DNS. This is not necessary, as far as I understand the manual, but however... On those other addresses, the machines share NFS, and I want to replace this NFS share with the OCFS2 volume. The addresses did not work either. Sunil, thanks for the help! My OCFS2 cluster is working (now ;-)) Regards, Werner> > Werner Flamme wrote: >> Hi, >> >> I want to share a partition of an external disk between 3 machines. >> It is located on a Sun StoreTek 3510 and bound (via multipathing) >> to all 3 machines. >> >> One machine (say mach1) is the node I used to create the filesystem >> and setup /etc/ocfs2/cluster.conf. I copied the file to the other 2 >> machines. All use the same /etc/multipath.conf[1], too. All start >> /etc/init.d/ocfs2 at boot, all are running with the same OS and >> patches (SLES 9 SP 4) and the same version of ocfs2 tools >> (ocfs2-tools-1.2.1-4.2, ocfs2console-1.2.1-4.2). The volume is >> mounted to /zdisk/ocfs2 on all machines (theoretically). >> >> When I mount the volume on mach1, it is successful. On the other >> machines (mach2 and mach3), I get: "mount.ocfs2: Transport endpoint >> is not connected while mounting /dev/disk/by-name/commondisk-part1 >> on /zdisk/ocfs2". On mach1, mount shows "/dev/dm-4 on /zdisk/ocfs2 >> type ocfs2 (rw,_netdev,heartbeat=local)". >> >> [1] /etc/multipath.conf translates the WWID into >> /dev/disk/by-name/commondisk-part1 >> >> Following my /etc/ocfs2/cluster.conf: node: ip_port = 7777 >> ip_address = 141.65.128.114 number = 0 name = mach1 cluster = ocfs2 >> >> >> node: ip_port = 7777 ip_address = 141.65.128.113 number = 1 name >> mach2 cluster = ocfs2 >> >> node: ip_port = 7777 ip_address = 141.65.128.112 number = 2 name >> mach3 cluster = ocfs2 >> >> cluster: node_count = 3 name = ocfs2 >> >> What do I do wrong? How may I succes in mounting the volume on all >> 3 machines? >> >> Regards, Werner >> >> > >-- Werner Flamme, Abt. WKDV Helmholtz-Zentrum f?r Umweltforschung GmbH - UFZ Permoserstr. 15 - 04318 Leipzig Tel.: (0341) 235-1921 - Fax (0341) 235-451921 http://www.ufz.de - eMail: werner.flamme at ufz.de