EVA8100 failover/failback

[Rudolf Gabler]

Hi,

the storage for our rx2660 running SLES10 is grouped as follows:

pool a: Controller A failover/failback
pool b: Controller B failover/failback

on our EVA 8100. On the rx2660 the pools are added by striping to a lvm volume. Through the hp_fibre software I see (failover setting) with adapter_info -L that all luns are on one (of the two HBAs) which is not load balanced but limited by the throughput of one HBA.

How can I achieve equal balancing of the luns through both HBAs with respect to a failover scenario?

Best regards,
Rudi Gabler

[Fernando Cabal]

Hello

the EVA is not an active/active SAN so the config you want to make is not possible.

Cheers!

[Fernando Cabal]

Rudi

I recently completed a deployment on an EVA8xxx and EVA4xxx and I learned that the newer products do support failover.

This is the setup for SLES 10 SP1 or SP2, using SLES drivers provided by Novell.


modprobe.conf.local:
options qla2xxx ql2xmaxqdepth=16 qlport_down_retry=10 ql2xloginretrycount=30 ql2xfailover=0

(ql2xfailover=0 is the real option that turns of failover on the Qlogic or HP Qlogic based cards and allos to use the 2 paths)



/etc/multipath.conf:
defaults {
user_friendly_names yes
multipath_tool "/sbin/multipath -v0"
udev_dir /dev
polling_interval 10
default_selector "round-robin 0"
default_path_grouping_policy failover
default_getuid_callout "/sbin/scsi_id -g -u -s /block/%n"
default_prio_callout "/bin/true"
rr_min_io 100
failback immediate
default_features "0"
}

blacklist {
devnode "^(ram|raw|loop|fd|md|dm-|sr|scd|st)*"
devnode "^hd*]"
devnode "^cciss!cd**]"
}

devices {
device {
vendor "(COMPAQ|HP)"
product "HSV11|HSV20|HSV300"
getuid_callout "/sbin/scsi_id -g -u -s /block/%n"
features "0"
hardware_handler "0"
path_grouping_policy group_by_prio
failback immediate
rr_weight uniform
no_path_retry 12
rr_min_io 100
path_checker tur

}
}
## this is to obtain names that don-t change after reeboot; name to be used safely in /etc/fstab
multipaths {
multipath {
wwid 000000000000000000000000
alias mpath1
}
}


Cheers
Fernando

[Rudolf Gabler]

Hi Fernando,

to your first reply:
the eva8100 is an active/active by default. Older Storage (i.e.eva5000 are active/active by the latest firmware).

to your second reply:

failover is supported since years with the hp (qlogic) failover drivers. On the other hand, these drivers also solve the multipath problem. I don't understand why one should turn the failover support down. This is a wanted property.

[Uwe Zessin]

If you want to stay with the QLogic failover, the "ql2xlbType" should allow a simple round-robin load balancing of virtual disks across paths. But I doubt that the driver is so clever and understands optimized and non-optimized (through the non-owning controller) paths. Depending on your SAN topology, the setting might not accomplish what you want.

I've never tried it, but maybe you can do explicit path selection with the SANsurfer HBA software (optinal install).


> I don't understand why one should turn the failover support down.

In Fernando's setup, the multipath failover is done by the devicemapper software on top of the QLogic adapter driver, so the driver must not filter any paths.

If you want to stay with the QLogic failover, of course, you should not turn off its multipath failover handling.