View Full Version : Enabling multipath for boot volume SLES 11 SP3 and EMC VNX

16-Jul-2014, 15:33
I am working on a project to document how to connect SAN disk to a Linux on System z server and need some help.

The SAN team provided me with four 50GB LUNs on an EMC VNX that appear to linux as vendor=DGC and product=VRAID as well as two 100GB LUNs on an IBM V7000.

The Enterprise Server is a z114, running z/VM 6.2, with a 4 port fcp card which is connected to two Brocade switches.

Phase 1 of my project is to add SAN disk to an existing SLES 11 SP3 system that is installed and running on ECKD dasd (from a guest under z/VM 6.2).

Phase 2 of my project is to install SLES 11 SP3 onto SAN disk (from a guest under z/VM 6.2).

Phase 1 was successful as I was able to access, with multipath enabled, all 6 of the LUNs.

FYI - Though the SLES 11 doc said that the EMC VNX was supported out of the box I was not able to enable multipath without extensive modification of /etc/multipath.conf. First off, the default entry for this hardware in the sample file /usr/share/doc/packages/multipath-tools/multipath.conf.defaults is incorrect (at least in my environment). I copied the following code to /etc/multipath.conf and uncommented each line. (Note: I had already copied /usr/share/doc/packages/multipath-tools/multipath.conf.synthetic to /etc/multipath.conf.)

# device {
# vendor "DGC"
# product ".*"
# product_blacklist "LUNZ"
# path_grouping_policy "group_by_prio"
# path_checker "emc_clariion"
# features "1 queue_if_no_path"
# hardware_handler "1 emc"
# prio "emc"
# failback "immediate"
# rr_weight "uniform"
# no_path_retry 60
# }

In order to get multipath enabled on the EMC VNX, I also had to change path_checker to "readsector0", hardware_handler to "1 alua", and prio to "alua". The IBM LUNs were recognized without having to add or modify a device stanza.

I'm having trouble with Phase 2 in getting multipath enabled on the EMC VNX. I haven't attempted to install SLES 11 on the IBM device yet though I expect that it will work as per the SLES documentation (Storage manual). I have successfully installed SLES 11 onto one of the EMC VNX LUNs and have been able to shutdown the guest and IPL from the SAN disk. I have created /etc/multipath.conf with similar parameters as I used for phase 1, along with other parameter recommendations from the Storage manual.

The /etc/multipath.conf file I'm using is below.

## This is a template multipath-tools configuration file
## Uncomment the lines relevent to your environment
defaults {
polling_interval 30
path_checker tur
no_path_retry 5
user_friendly_names yes
dev_loss_tmo 90
fast_io_fail_tmo 5
blacklist {
devnode "^dasd[a-z][[0-9]*]"
device {
vendor "DGC"
product "LUNZ"
device {
vendor "EMC"
product "LUNZ"
multipaths {
multipath {
wwid 3600601608fa034006e987a05a0dfe311
alias EMC-VNX_LUN0
devices {
device {
vendor "DGC"
product "*"
product_blacklist "LUNZ"
path_grouping_policy "group_by_prio"
path_checker "readsector0"
features "1 queue_if_no_path"
hardware_handler "1 alua"
prio "alua"
failback "immediate"
rr_weight "uniform"
no_path_retry 60

I have followed the procedure in the manual to add 'dm-multipath' to the INITRD_MODULES= parameter in /etc/sysconfig/kernel, save the file, run mkinitrd, run zipl, add boot.multipath and multipathd to the startup processes, then shutdown and IPL. When I attempt to IPL the server, linux begins to access the SAN disk but when it starts to enable multipath it displays about 30 lines of error-type messages that repeat until I terminate the boot (via z/VM command #CP I CMS).

Any thoughts on what else I need to do enable multipath on the boot volume? I need the path redundancy in case we decide to use SAN disk for some or all of our Enterprise servers.

Note: the attached zip file contains the Word document for this phase of the project.


26-Aug-2014, 13:54

I was able to get multipathing to work on the Linux installation to the EMC VNX SAN disc. It wasn enabled automatically by the installation routines. The SAN team modified a setting in the EMC box (they didn't tell me what they changed) and asked me to reinstall SLES 11 SP3. When Linux booted at the end of the install, multipathing was enabled.

I also installed Linux onto an IBM SAN disc and multipathing was enabled by the installation routines there as well.

A bug report was opened for the syntax error in /usr/share/doc/packages/multipath-tools/multipath.conf.defaults but I haven't heard of when a fix will be issued.