Hi ;

I have been studying on a problem and could not find the answer ;
here is shortly the problem and my struggle

there are 2 unisys es7000 itanium servers running redhat enterprise linux 4 up 4 ;
8 gb ram, 70 gb hdd and 2 hba's each ;
There will be oracle rac on these systems ; so a shared disk storage EMC CX500 is attached them through 2 hba's to support failover.(1 tb + 250 gb storage)
after installing operating systems
node1 fdisk -l

and node 2 fdisk -l is
All the hardare in server are indentical ; only serial number ;
My question is what happens to node 2 ; where is /dev/sdb ?
Why did it just bypass that letter and uses next ?

I have read some manuals , man pages but none of them helped;
I tried to create a special device file name "emcpower" by writing custom rules in /etc/udev/rules.d/ that worked
I used uuid to write special rules
I was successful in mounting by "mount /dev/emcpower /mnt/myhugedrive" then I wrote some files to it.but when I created partitions like 500gb + 500gb in my disk it failed so I was failed to use special rules by using uuid .I changed NAME="emcpower" --> NAME="emcpower%e" to use them just as
/dev/sda --> /dev/emcpower
/dev/sda1 --> /dev/emcpower1
but no worked
Is there a way to bind devices to specific device files ?

Using e2label was the fisrt thing I thougt but it is valid only using ext2 ext3 file systems as far as ı know ; I plan to use ocfs file system on that disk because of oracle rac;

PS: some hardware info
lspci (hba cards info)
2d:01.0 Fibre Channel: Emulex Corporation Thor LightPulse Fibre Channel Host Adapter (rev 01)
37:01.0 Fibre Channel: Emulex Corporation Thor LightPulse Fibre Channel Host Adapter (rev 01)

[root@node1]# cat /proc/meminfo
MemTotal: 8204496 kB

/dev/sda's partition table canot be seen via fdisk ; parted will print it;


Any help will be appreciated;
Thanks ;
Have a nice happy valentine's day if even I will not

I have been looking for reason but no way ?
Am I the only one facing this problem in te world ?

There's got to be a way I would think, because Veritas Volume manager and EMC Powerpath both essentially just map the raw device to a named device. And if it helps, I've noticed that all of my servers with multipath LUNS see the same lun twice. So /dev/sda and /dev/sdc might be the same lun, just different paths.

Well the problem seems to be a little tricky, but lets take it one by one.

Firstly, you created the partitions using fdisk and then tried to bind the device. But here is a catch, when you do this on a single node then the other node would not be syncing the partition table. So firstly, on the second machine (where you DID NOT do fdisk to create the partition) open the disk using fdisk command and without changing the partition table. simply do a "w". This will ensure that the shared disk syncs on both the nodes of the cluster.

Secondly, Since you are using a device file to attach to sd* series. Better create it as a character device.
like:-

mknod /dev/emc** c 162 **

This way you would be able to use it as a raw partition. If you are planning to use the OCFS then tell me what exactly do you get when you do:-

fdisk /dev/sdb

-Rahul.

thank cambie and rahulk ;

Firstly as cambie has mentioned "the sdb and sdc may be the same phisical partition "
that was true before we unplugged the second fiber cables;
And the size were identical for the multi pathed devices; but as you can see above they all have different sizes;

because emc has not a supported driver for rhel4 up4 64bit on itanıum ;
and we all have these problems ; we unplugged the secondary cables for the risk of mounting same phisical disk in different paths manually by unexperienced customer

Whatelse if emc can deliver the driver perhaps our problems will end ;


Secondary;
partprobe reads partition tables for disks which are created by fdisk or parted;
and if ı change partition table from one node ; second can access the partitions after partprobed;

I will (even am using ; because one node is in production now) use ocfs ;
but fdisk /dev/sdb gives the same output above;

again very much thank you;

I got similar issue on my SLES 10 machine, it is happened after upgrade the SLES 10 SP2 to SLES 10 SP3 P5. we found the kernel is not comparability to PP version installed before. after the PP version upgrade into 5.5.0, we still not able to present the partition on my machine. everyone can help to recover the disk partition without destroyed the Data from the LUN assigned.


Disk /dev/sda: 1717.9 GB, 1717986918400 bytes
255 heads, 63 sectors/track, 208866 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes

Disk /dev/sda doesn't contain a valid partition table

Disk /dev/sdb: 1717.9 GB, 1717986918400 bytes
255 heads, 63 sectors/track, 208866 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes

Disk /dev/sdb doesn't contain a valid partition table

Disk /dev/sdc: 1717.9 GB, 1717986918400 bytes
255 heads, 63 sectors/track, 208866 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes

Disk /dev/sdc doesn't contain a valid partition table

Disk /dev/sdd: 1717.9 GB, 1717986918400 bytes
255 heads, 63 sectors/track, 208866 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes

Disk /dev/sdd doesn't contain a valid partition table