Using different configuration for arrays with identical vendor/product

Using different configuration for arrays with identical vendor/product

[Tore Anderson]

Hey.  I've got an EMC CX200 and a CX3-40 in my SAN.  The latter one
supports ALUA, which I'm using as much as I can, the former does not.
The hosts need to be able to access volumes on both the arrays at the
same time.

Unfortunately the volumes on both of these boxes have vendor «DGC» and
product «RAID 5», so making two different device sections for them with
different hardware_handler and path_checker settings isn't trivial.  I
figured I could simply override the two settings in the multipath{}
blocks for volumes on the CX200, making the device{} block specific for
the CX3-40.  Doesn't appear to work, though, the path_checker- and
hardware_handler-settings in the multipath{} blocks appears to be
silently ignored.

Am I out of luck or is there any other way to do what I want?

For reference, this is the device section I use for the CX200 in
standard «trespassing» mode:

device {
  vendor               DGC.*
  product              RAID.*
  hardware_handler     "1 emc"
  path_checker         emc_clariion
  prio_callout         "/sbin/mpath_prio_emc /dev/%n"
  features             "1 queue_if_no_path"
  path_grouping_policy group_by_prio
  failback             immediate
}

And for the CX3-40 in ALUA mode:

device {
  vendor               DGC.*
  product              RAID.*
  hardware_handler     "0"
  path_checker         tur
  prio_callout         "/sbin/mpath_prio_emc /dev/%n"
  features             "1 queue_if_no_path"
  path_grouping_policy group_by_prio
  failback             immediate
}

These both work fine on hosts that only connect to one of the arrays at
a time.

Regards
-- 
Tore Anderson

RE: Using different configuration for arrays with identicalvendor/product

[berthiaume_wayne]

Hi Tore.

	You should be able to use the emc_handler for both arrays.
Depending on the distro you are using only implicit ALUA is supported,
explicit ALUA is in being worked on for upcoming distros. Whether the
storage group within the array attached to the server is ALUA or not is
determined by the failover mode you set for the host on that array. At
the server side you will not know the difference between the two
failover modes, PNR or ALUA.

Regards,
Wayne.
EMC Corp 

-----Original Message-----
From: dm-devel-bounces@redhat.com [mailto:dm-devel-bounces@redhat.com]
On Behalf Of Tore Anderson
Sent: Friday, March 28, 2008 7:24 AM
To: dm-devel@redhat.com
Subject: [dm-devel] Using different configuration for arrays with
identicalvendor/product

Hey.  I've got an EMC CX200 and a CX3-40 in my SAN.  The latter one
supports ALUA, which I'm using as much as I can, the former does not.
The hosts need to be able to access volumes on both the arrays at the
same time.

Unfortunately the volumes on both of these boxes have vendor <DGC> and
product <RAID 5>, so making two different device sections for them with
different hardware_handler and path_checker settings isn't trivial.  I
figured I could simply override the two settings in the multipath{}
blocks for volumes on the CX200, making the device{} block specific for
the CX3-40.  Doesn't appear to work, though, the path_checker- and
hardware_handler-settings in the multipath{} blocks appears to be
silently ignored.

Am I out of luck or is there any other way to do what I want?

For reference, this is the device section I use for the CX200 in
standard <trespassing> mode:

device {
  vendor               DGC.*
  product              RAID.*
  hardware_handler     "1 emc"
  path_checker         emc_clariion
  prio_callout         "/sbin/mpath_prio_emc /dev/%n"
  features             "1 queue_if_no_path"
  path_grouping_policy group_by_prio
  failback             immediate
}

And for the CX3-40 in ALUA mode:

device {
  vendor               DGC.*
  product              RAID.*
  hardware_handler     "0"
  path_checker         tur
  prio_callout         "/sbin/mpath_prio_emc /dev/%n"
  features             "1 queue_if_no_path"
  path_grouping_policy group_by_prio
  failback             immediate
}

These both work fine on hosts that only connect to one of the arrays at
a time.

Regards
-- 
Tore Anderson

--
dm-devel mailing list
dm-devel@redhat.com
https://www.redhat.com/mailman/listinfo/dm-devel

Re: Using different configuration for arrays with identicalvendor/product

[Tore Anderson]

* berthiaume_wayne@emc.com

> 	You should be able to use the emc_handler for both arrays.
> Depending on the distro you are using only implicit ALUA is supported,
> explicit ALUA is in being worked on for upcoming distros. Whether the
> storage group within the array attached to the server is ALUA or not is
> determined by the failover mode you set for the host on that array. At
> the server side you will not know the difference between the two
> failover modes, PNR or ALUA.

Hi Wayne, thanks for your reply.

So I can use the emc_clariion hardware handler towards an array
configured for ALUA mode.  Wasn't aware of that.  I feel it's better to
let the array handle the trespassing on its own based on the ratio of
«non-optimal» I/O operations (using hardware_handler 0), but I can live
with host-initiated trespasses on PG swithces/inits.

Anyway - there's still the problem of the path_checker.  If I'm using
«emc_clariion» for the array configured for ALUA mode (CLARiiON Open,
mode 4), all paths is perceived to be failed.  If I'm using «tur» for
the array configured for PNR mode (CLARiiON Open, mode 1), the passive
path is perceived to be failed.

Hence, if I use «emc_clariion» I won't be able to access the volumes on
the CX3-40 at all, while if I use «tur» dm-multipath won't switch PGs
when there's a (real) failure on the active path to the CX200.  None of
these are acceptable alternatives for production use, so the only
workaround I've found so far is to configure the CX3-40 to use PNR mode
(which sucks since ALUA is way better).  Do you have another trick up
your sleeve for me?

Regards
-- 
Tore Anderson