Powerpath Path Failover

Hello Sandy,

Is PowerPath licensed ? What array is this system connected to ? If it's clariion..what failover mode is set to ?

what cluster s/w? Red hat cluster Suite( which version) with Quorum disk?

what you have in /etc/modprobe.conf? We have the following entry which disables the fail over at HBA level rather PP/Array handles fail over. ( that is what i remember about this entry)

options qla2xxx ConfigRequired=0 ql2xfailover=0

Hi Dynamox,

Yes the powerpath is licensed, and it is connected to Clariion box, I give some output from powermt display as below:

Pseudo name=emcpowerb
CLARiiON ID=XXXX [SERVER1_Clust]
Logical device ID=60060160C9CF1A00EEB33F4C078BDE11 [LUN 68]
state=alive; policy=CLAROpt; priority=0; queued-IOs=0
Owner: default=SP A, current=SP A Array failover mode: 1
==============================================================================
---------------- Host --------------- - Stor - -- I/O Path - -- Stats ---
### HW Path I/O Paths Interf. Mode State Q-IOs Errors
==============================================================================
3 qla2xxx sdb SP A2 active alive 0 0
3 qla2xxx sde SP B2 active alive 0 0
4 qla2xxx sdh SP A3 active alive 0 0
4 qla2xxx sdk SP B3 active alive 0 0

Hi SKT,

Yes its a redhat cluster suite with a quorum disk. I'm not sure what version is installed, how can i check that?

I have read some articles about the /etc/modprobe.conf options, but when i add those lines, i get error messages whenever i do a modprobe. Some said have to do mkinitrd after i add the options, but whenever i do mkinitrd, it saying that there is no changes on the kernel...

Can you enlighten me

Regards,
Sandy

Post `cman_tool status`

cat /etc/modprobe.conf. I can compare with mine. You dont need to do any modprobe after the change. complete mkinitrd and reboot.

DId u check the size(or do a diff between old.img and new.img) of the *.img file after mkinitrd even though it reports no change in the kernel.

also post /`cat proc/scsi/qla2xxx/3`

Hi SKT, thanks for your replies

[root@SERVER1 ~]# cat /etc/modprobe.conf
alias eth0 e1000e
alias eth1 e1000e
alias eth2 bnx2
alias eth3 bnx2
alias scsi_hostadapter mptbase
alias scsi_hostadapter1 mptsas
alias scsi_hostadapter2 ata_piix
alias scsi_hostadapter3 qla2xxx
alias bond0 bonding
options qla2xxx ql2xfailover=0 ConfigRequired=0
# options bond0 mode=balance-alb miinom=100
###BEGINPP
include /etc/modprobe.conf.pp
###ENDPP

[root@SERVER1 ~]# cman_tool status
Version: 6.1.0
Config Version: 11
Cluster Name: SERVER1CLUST
Cluster Id: 24958
Cluster Member: Yes
Cluster Generation: 220344
Membership state: Cluster-Member
Nodes: 2
Expected votes: 2
Quorum device votes: 1
Total votes: 3
Quorum: 2
Active subsystems: 9
Flags: Dirty
Ports Bound: 0 177
Node name: SERVERC01-PRIV
Node ID: 1
Multicast addresses: 239.192.97.223
Node addresses: 172.21.62.11

for the mkinitrd, its not even creating a new img file, let me try to do that again once again and post the result later on,

I cannot find qla2xxx under the folder /proc/scsi

Regards,
Sandy

Here is the one to over write(force) the existing file. Take a backup first

#cd /boot
#mkinitrd -f initrd-$(uname -r).img $(uname -r)

Hi,

Yes i use the -f options when running the mkinitrd command, i forgot the exact error message. when i can down the cluster, i'll try to do it again and post the result.

Regards,
Sandy

Hi,

Please find as below when i change the /etc/modprobe.conf and then do mkinitrd. I got message for 'No modules available for kernel "initrd-2.6.18-128.el5"'

[root@SERVER1 boot]# cat /etc/modprobe.conf
alias eth0 e1000e
alias eth1 e1000e
alias eth2 bnx2
alias eth3 bnx2
alias scsi_hostadapter mptbase
alias scsi_hostadapter1 mptsas
alias scsi_hostadapter2 ata_piix
alias scsi_hostadapter3 qla2xxx
alias bond0 bonding
options qla2xxx ql2xfailover=0 ConfigRequired=0
###BEGINPP
include /etc/modprobe.conf.pp
###ENDPP

[root@SERVER1 boot]# ll
total 12474
-rw-r--r-- 1 root root 64994 Dec 17 2008 config-2.6.18-128.el5
drwxr-xr-x 2 root root 1024 Aug 14 10:49 grub
-rw------- 1 root root 3230851 Aug 14 10:16 initrd-2.6.18-128.el5.img
-rw------- 1 root root 3230851 Sep 20 11:13 initrd-2.6.18-128.el5.img.backup
-rw-r--r-- 1 root root 2982241 Aug 15 02:57 initrd-2.6.18-128.el5kdump.img
drwx------ 2 root root 12288 Aug 14 10:11 lost+found
-rw-r--r-- 1 root root 102182 Dec 17 2008 symvers-2.6.18-128.el5.gz
-rw-r--r-- 1 root root 1188481 Dec 17 2008 System.map-2.6.18-128.el5
-rw-r--r-- 1 root root 1889308 Dec 17 2008 vmlinuz-2.6.18-128.el5

[root@SERVER1 boot]# mkinitrd -f initrd-2.6.18-128.el5.img initrd-2.6.18-1 28.el5
No modules available for kernel "initrd-2.6.18-128.el5".

[root@SERVER1 boot]# ll
total 12474
-rw-r--r-- 1 root root 64994 Dec 17 2008 config-2.6.18-128.el5
drwxr-xr-x 2 root root 1024 Aug 14 10:49 grub
-rw------- 1 root root 3230851 Aug 14 10:16 initrd-2.6.18-128.el5.img
-rw------- 1 root root 3230851 Sep 20 11:13 initrd-2.6.18-128.el5.img.backup
-rw-r--r-- 1 root root 2982241 Aug 15 02:57 initrd-2.6.18-128.el5kdump.img
drwx------ 2 root root 12288 Aug 14 10:11 lost+found
-rw-r--r-- 1 root root 102182 Dec 17 2008 symvers-2.6.18-128.el5.gz
-rw-r--r-- 1 root root 1188481 Dec 17 2008 System.map-2.6.18-128.el5
-rw-r--r-- 1 root root 1889308 Dec 17 2008 vmlinuz-2.6.18-128.el5

Shouldn't the command be

mkinitrd -v initrd-2.6.18-128.el5.img initrd-2.6.18-1 28.el5

adding -v would give verbose info; U will get a lot info on the stadard o/p which may help to know what is going on.

try the below and make sure the modules for current kernels are available.
#cd /lib/modules/`uname -r`

I think these created when respcetive kernel-devel/kernel-headers rpms are insatlled. can u verify if they are installed.

Also you dont need any outage to run the mkinitrd, rather put/create the img file on a different folder.

Hi All,

I do some additional testing, it appears that when I'm not using cluster to mount the filesystem, the powerpath is working just fine. Whenever I unplug 1 path, the mount point will still be accessible.

But when I bring up the cluster and unplug 1 path, it will make my cluster failover to the other node. And to make it more interesting, if I manually swing it back to the node with 1 path unplugged, also no problem.

Any clues?

Regards,
Sandy

I've seen the same behavior on a competitor's system with a Windows cluster where 1 path to 1 of the 2 hosts failed and the cluster failed over to the other node. We're still not sure what caused this.

I see the note "I cannot find qla2xxx under the folder /proc/scsi".
In all my linux systems with qla/PP i can see them. Normally i see them disappearing them after a kernel upgrade and driver installation fixes them.

can u try reinstallng the qlogic driver.?

Ther driver version we use is

QLogic PCI to Fibre Channel Host Adapter for QLA2460:
Firmware version 4.00.26 [IP] , Driver version 8.01.07.15
ISP: ISP2422

Hi All,

I try to dig on the cluster side, I found out that linux cluster has the timeout/TKO for the quorum disk. I changed this value to be above 30 sec and it is all working as expected now.

Thank you all for the help

Regards
Sandy

u mena previos value was 30. ? We have the tko set to 10