Clariion CX600 automatic failback support

Clariion CX600 automatic failback support

[Phil Lowden (plowden)]

I'm using LVM2 with device-mapper-multipath 0.4.5 on GA RHEL 4 update 2 
release, kernel version 2.6.9-22.ELsmp.  Storage is 4 Clariion CX600
LUNs, 
2 with a primary path on SP A and 2 on SP B.  HBAs are QLA3240 with
firmware 3.03.15 IPX and the RedHat-distributed driver version
8.01.00b5-rh2.

When I disrupt one path by disabling a host or SP switch port,
failover works great but failback doesn't happen automatically
with the current config (below).  By this I mean
when the connection to e.g. SP B is restored, all
4 LUNs stay trespassed to SP A.  Is this by design?
Or is there support for automatic failback, i.e.
I/O is paused and a trespass is issued to restore
the 2 SP B LUNs to their primary paths?

Of course manual failback is possible, but without
quiescing I/O I found I was able to munge my LVM2
objects up quite nicely.  Manual failback with quiesced 
I/O is fine.

Here's the /etc/multipath.conf:

defaults {
       multipath_tool  "/sbin/multipath -v0"
       udev_dir        /dev
       polling_interval 10
       default_selector        "round-robin 0"
       default_path_grouping_policy    multibus
       default_getuid_callout  "/sbin/scsi_id -g -u -s /block/%n"
       default_prio_callout    "/bin/true"
       default_features        "0"
       rr_wmin_io              100
       failback                immediate
}
devices {
    device {
        vendor "DGC     "
        product "RAID 5        "
        #path_grouping_policy group_by_serial
        path_grouping_policy failover
        getuid_callout "/sbin/scsi_id -g -u -s /block/%n"
        path_checker    emc_clariion
        path_selector   "round-robin 0"
        features        "0"
        hardware_handler        "1 emc"
    }
}

Here's the multipath -l output (with both paths up)
and a few commands to help make sense of it:

[root@sandbox ~]# multipath -l
3600601604b600d00743e69d8862fda11
[size=1 GB][features="0"][hwhandler="1 emc"]
\_ round-robin 0 [active]
  \_ 1:0:0:3 sdd  8:48    [active]
\_ round-robin 0 [enabled]
  \_ 0:0:0:3 sdh  8:112   [active]

3600601604b600d00773e69d8862fda11
[size=4 GB][features="0"][hwhandler="1 emc"]
\_ round-robin 0 [active]
  \_ 1:0:0:0 sda  8:0     [active]
\_ round-robin 0 [enabled]
  \_ 0:0:0:0 sde  8:64    [active]

3600601604b600d00763e69d8862fda11
[size=3 GB][features="0"][hwhandler="1 emc"]
\_ round-robin 0 [active]
  \_ 1:0:0:1 sdb  8:16    [active]
\_ round-robin 0 [enabled]
  \_ 0:0:0:1 sdf  8:80    [active]

3600601604b600d00753e69d8862fda11
[size=2 GB][features="0"][hwhandler="1 emc"]
\_ round-robin 0 [active]
  \_ 1:0:0:2 sdc  8:32    [active]
\_ round-robin 0 [enabled]
  \_ 0:0:0:2 sdg  8:96    [active]

# ls -l /dev/mpath
total 0
lrwxrwxrwx  1 root root 7 Oct 21 12:56 3600601604b600d00743e69d8862fda11
-> ../dm-3
lrwxrwxrwx  1 root root 7 Oct 21 12:56 3600601604b600d00753e69d8862fda11
-> ../dm-2
lrwxrwxrwx  1 root root 7 Oct 21 12:56 3600601604b600d00763e69d8862fda11
-> ../dm-1
lrwxrwxrwx  1 root root 7 Oct 21 12:56 3600601604b600d00773e69d8862fda11
-> ../dm-0

# pvs
  PV         VG      Fmt  Attr PSize    PFree  
  /dev/dm-0  vgtest2 lvm2 a-      4.00G   3.90G
  /dev/dm-1  vgtest2 lvm2 a-      3.00G   2.90G
  /dev/dm-2  vgtest2 lvm2 a-      2.00G   1.90G
  /dev/dm-3  vgtest2 lvm2 a-   1020.00M 920.00M

Thanks in advance,
Phil Lowden

Re: Clariion CX600 automatic failback support

[Brian Long]

On Wed, 2005-10-26 at 15:37 -0400, Phil Lowden (plowden) wrote:
> I'm using LVM2 with device-mapper-multipath 0.4.5 on GA RHEL 4 update 2 
> release, kernel version 2.6.9-22.ELsmp.  Storage is 4 Clariion CX600
> LUNs, 
> 2 with a primary path on SP A and 2 on SP B.  HBAs are QLA3240 with
> firmware 3.03.15 IPX and the RedHat-distributed driver version
> 8.01.00b5-rh2.
> 
> When I disrupt one path by disabling a host or SP switch port,
> failover works great but failback doesn't happen automatically
> with the current config (below).  By this I mean
> when the connection to e.g. SP B is restored, all
> 4 LUNs stay trespassed to SP A.  Is this by design?
> Or is there support for automatic failback, i.e.
> I/O is paused and a trespass is issued to restore
> the 2 SP B LUNs to their primary paths?
> 
> Of course manual failback is possible, but without
> quiescing I/O I found I was able to munge my LVM2
> objects up quite nicely.  Manual failback with quiesced 
> I/O is fine.

No one has automatic failback working on Clarrion storage?

/Brian/
-- 
       Brian Long                      |         |           |
       IT Data Center Systems          |       .|||.       .|||.
       Cisco Linux Developer           |   ..:|||||||:...:|||||||:..
       Phone: (919) 392-7363           |   C i s c o   S y s t e m s

Re: Clariion CX600 automatic failback support

[Bernd Zeimetz]

Hi,

> No one has automatic failback working on Clarrion storage?

failback seems to work with the latest multipath-tools, but since evms (with 
lvm-plugin..) doesn't recognize the multipath-devices anymore (it uses only 
the blank sd* devices) I never had a real chance to test it - see my mail to 
the lsit from saturday.


Bernd

Re: Clariion CX600 automatic failback support

[Christophe Varoqui]

> 
> No one has automatic failback working on Clarrion storage?
> 
Do you reproduce with upstream ?

Regards,
cvaroqui

Re: Clariion CX600 automatic failback support

[Brian Long]

On Mon, 2005-10-31 at 15:07 +0100, Christophe Varoqui wrote:
> > 
> > No one has automatic failback working on Clarrion storage?
> > 
> Do you reproduce with upstream ?

No.  I imagine we will log a support request with Red Hat to get it
working on RHEL 4 U2.  If they want to give us an updated device-mapper-
multipath RPM to try to reproduce with upstream, that's up to them.
Phil and I were just wondering if there is a specific configuration
parameter to get the trespass to happen again once all paths are back
online.

/Brian/

-- 
       Brian Long                      |         |           |
       IT Data Center Systems          |       .|||.       .|||.
       Cisco Linux Developer           |   ..:|||||||:...:|||||||:..
       Phone: (919) 392-7363           |   C i s c o   S y s t e m s

Re: Clariion CX600 automatic failback support

[Christophe Varoqui]

On Mon, Oct 31, 2005 at 10:51:38AM -0500, Brian Long wrote:
> On Mon, 2005-10-31 at 15:07 +0100, Christophe Varoqui wrote:
> > > 
> > > No one has automatic failback working on Clarrion storage?
> > > 
> > Do you reproduce with upstream ?
> 
> No.  I imagine we will log a support request with Red Hat to get it
> working on RHEL 4 U2.  If they want to give us an updated device-mapper-
> multipath RPM to try to reproduce with upstream, that's up to them.
> Phil and I were just wondering if there is a specific configuration
> parameter to get the trespass to happen again once all paths are back
> online.
> 
Every check interval, the daemon re-evaluates path priorities (through
external prioritizers), then re-compute path-groups priorities. If PG
priority order change, the highest-prio PG might be activated :

- immediatly, if "failback" is set to immediate
- never, if "failback" is set to manual
- in a number of prio-order stable consecutive checks, if failback
  set to #

Thus, failback/no-failback is a property of the external prioritizer, plus
a framework policy switch in the name of the "failback" config file keyword.

Well, at least it is the current design upstream.
You'll have to dig the changelog to see if RedHat packages all the
necassary bits.

Regards,
cvaroqui

RE: Clariion CX600 automatic failback support

[goggin, edward]

On Date: Mon, 31 Oct 2005 08:58:06 -0500
Brian Long <brilong@cisco.com> wrote 

> 
> On Wed, 2005-10-26 at 15:37 -0400, Phil Lowden (plowden) wrote:
> > I'm using LVM2 with device-mapper-multipath 0.4.5 on GA 
> RHEL 4 update 2 
> > release, kernel version 2.6.9-22.ELsmp.  Storage is 4 Clariion CX600
> > LUNs, 
> > 2 with a primary path on SP A and 2 on SP B.  HBAs are QLA3240 with
> > firmware 3.03.15 IPX and the RedHat-distributed driver version
> > 8.01.00b5-rh2.
> > 
> > When I disrupt one path by disabling a host or SP switch port,
> > failover works great but failback doesn't happen automatically
> > with the current config (below).  By this I mean
> > when the connection to e.g. SP B is restored, all
> > 4 LUNs stay trespassed to SP A.  Is this by design?
> > Or is there support for automatic failback, i.e.
> > I/O is paused and a trespass is issued to restore
> > the 2 SP B LUNs to their primary paths?
> > 
> > Of course manual failback is possible, but without
> > quiescing I/O I found I was able to munge my LVM2
> > objects up quite nicely.  Manual failback with quiesced 
> > I/O is fine.
> 
> No one has automatic failback working on Clarrion storage?

Unfortunately, this doesn't just work "out of the box" yet.

You need to setup a default failback setting of immediate in the
/etc/multipath.conf configuration file, ...

	defaults {
		failback 	immediate
	}

... or selectively set the failback setting to immediate in the
multipath config file for each EMC CLARiiON logical unit

	multipaths {
		multipath {
			wwid	360061 ...
			failback	immediate
		}
		...
	}

Setting this attribute value for the entire EMC CLARiiON class of
devices via the multipath config file "device" attribute doesn't
work because of a bug (mpp->hwe field is never set for multipathd)
in the multipathd code which is patched to work by the single line
patch below.  Due to the same bug, a patch I submitted many weeks
ago to libmultipath/hwtable.c to setup the default failback policy
for EMC CLARiiON to immediate also does not work.

Once the distributors pick up both this single line change and the
change in hwtable.c, immediate path group failback for CLARiiON
will just work.

-------------------------------------------------------------------
diff --git a/multipathd/main.c b/multipathd/main.c
--- a/multipathd/main.c
+++ b/multipathd/main.c
@@ -146,6 +146,7 @@ adopt_paths (struct vectors * vecs, stru
 		if (!strncmp(mpp->wwid, pp->wwid, WWID_SIZE)) {
 			condlog(4, "%s ownership set", pp->dev_t);
 			pp->mpp = mpp;
+			mpp->hwe = pp->hwe;
 		}
 	}
 }

Re: Clariion CX600 automatic failback support

[Christophe Varoqui]

> 
> Once the distributors pick up both this single line change and the
> change in hwtable.c, immediate path group failback for CLARiiON
> will just work.
> 
> -------------------------------------------------------------------
> diff --git a/multipathd/main.c b/multipathd/main.c
> --- a/multipathd/main.c
> +++ b/multipathd/main.c
> @@ -146,6 +146,7 @@ adopt_paths (struct vectors * vecs, stru
>  		if (!strncmp(mpp->wwid, pp->wwid, WWID_SIZE)) {
>  			condlog(4, "%s ownership set", pp->dev_t);
>  			pp->mpp = mpp;
> +			mpp->hwe = pp->hwe;
>  		}
>  	}
>  }
> 
Upstream has had a patch for that. It plug the mpp->hwe refresh in
setup_multipath().

Can you check it has no regression wrt your suggested fix ?

Regards,
cvaroqui

Re: Clariion CX600 automatic failback support

[Benjamin Marzinski]

On Wed, Nov 02, 2005 at 04:32:06PM -0500, goggin, edward wrote:
> On Date: Mon, 31 Oct 2005 08:58:06 -0500
> Brian Long <brilong@cisco.com> wrote 
> 
> > 
> > On Wed, 2005-10-26 at 15:37 -0400, Phil Lowden (plowden) wrote:
> > > I'm using LVM2 with device-mapper-multipath 0.4.5 on GA 
> > RHEL 4 update 2 
> > > release, kernel version 2.6.9-22.ELsmp.  Storage is 4 Clariion CX600
> > > LUNs, 
> > > 2 with a primary path on SP A and 2 on SP B.  HBAs are QLA3240 with
> > > firmware 3.03.15 IPX and the RedHat-distributed driver version
> > > 8.01.00b5-rh2.
> > > 
> > > When I disrupt one path by disabling a host or SP switch port,
> > > failover works great but failback doesn't happen automatically
> > > with the current config (below).  By this I mean
> > > when the connection to e.g. SP B is restored, all
> > > 4 LUNs stay trespassed to SP A.  Is this by design?
> > > Or is there support for automatic failback, i.e.
> > > I/O is paused and a trespass is issued to restore
> > > the 2 SP B LUNs to their primary paths?
> > > 
> > > Of course manual failback is possible, but without
> > > quiescing I/O I found I was able to munge my LVM2
> > > objects up quite nicely.  Manual failback with quiesced 
> > > I/O is fine.
> > 
> > No one has automatic failback working on Clarrion storage?
> 
> Unfortunately, this doesn't just work "out of the box" yet.
> 
> You need to setup a default failback setting of immediate in the
> /etc/multipath.conf configuration file, ...
> 
> 	defaults {
> 		failback 	immediate
> 	}
> 
> ... or selectively set the failback setting to immediate in the
> multipath config file for each EMC CLARiiON logical unit
> 
> 	multipaths {
> 		multipath {
> 			wwid	360061 ...
> 			failback	immediate
> 		}
> 		...
> 	}
> 
> Setting this attribute value for the entire EMC CLARiiON class of
> devices via the multipath config file "device" attribute doesn't
> work because of a bug (mpp->hwe field is never set for multipathd)
> in the multipathd code which is patched to work by the single line
> patch below.  Due to the same bug, a patch I submitted many weeks
> ago to libmultipath/hwtable.c to setup the default failback policy
> for EMC CLARiiON to immediate also does not work.
> 
> Once the distributors pick up both this single line change and the
> change in hwtable.c, immediate path group failback for CLARiiON
> will just work.

These changes are both in the RHEL4 U3 code.
 
> -------------------------------------------------------------------
> diff --git a/multipathd/main.c b/multipathd/main.c
> --- a/multipathd/main.c
> +++ b/multipathd/main.c
> @@ -146,6 +146,7 @@ adopt_paths (struct vectors * vecs, stru
>  		if (!strncmp(mpp->wwid, pp->wwid, WWID_SIZE)) {
>  			condlog(4, "%s ownership set", pp->dev_t);
>  			pp->mpp = mpp;
> +			mpp->hwe = pp->hwe;
>  		}
>  	}
>  }
> 
> --
> dm-devel mailing list
> dm-devel@redhat.com
> https://www.redhat.com/mailman/listinfo/dm-devel

RE: Clariion CX600 automatic failback support

[goggin, edward]

On Date: Thu, 3 Nov 2005 14:58:36 +0100
Christophe Varoqui <christophe.varoqui@free.fr> wrote

> > Once the distributors pick up both this single line change and the
> > change in hwtable.c, immediate path group failback for CLARiiON
> > will just work.
> > 
> > -------------------------------------------------------------------
> > diff --git a/multipathd/main.c b/multipathd/main.c
> > --- a/multipathd/main.c
> > +++ b/multipathd/main.c
> > @@ -146,6 +146,7 @@ adopt_paths (struct vectors * vecs, stru
> >  		if (!strncmp(mpp->wwid, pp->wwid, WWID_SIZE)) {
> >  			condlog(4, "%s ownership set", pp->dev_t);
> >  			pp->mpp = mpp;
> > +			mpp->hwe = pp->hwe;
> >  		}
> >  	}
> >  }
> > 
> Upstream has had a patch for that. It plug the mpp->hwe refresh in
> setup_multipath().
> 
> Can you check it has no regression wrt your suggested fix ?
> 
> Regards,
> cvaroqui

Looks good, except the call to extract_hwe_from_path in setup_multipath
must be before the call to select_pgfailback.

RE: Clariion CX600 automatic failback support

[Brian Long]

On Thu, 2005-11-03 at 13:16 -0500, goggin, edward wrote:
> On Date: Thu, 3 Nov 2005 14:58:36 +0100
> Christophe Varoqui <christophe.varoqui@free.fr> wrote
> 
> > > Once the distributors pick up both this single line change and the
> > > change in hwtable.c, immediate path group failback for CLARiiON
> > > will just work.
> > > 
> > > -------------------------------------------------------------------
> > > diff --git a/multipathd/main.c b/multipathd/main.c
> > > --- a/multipathd/main.c
> > > +++ b/multipathd/main.c
> > > @@ -146,6 +146,7 @@ adopt_paths (struct vectors * vecs, stru
> > >  		if (!strncmp(mpp->wwid, pp->wwid, WWID_SIZE)) {
> > >  			condlog(4, "%s ownership set", pp->dev_t);
> > >  			pp->mpp = mpp;
> > > +			mpp->hwe = pp->hwe;
> > >  		}
> > >  	}
> > >  }
> > > 
> > Upstream has had a patch for that. It plug the mpp->hwe refresh in
> > setup_multipath().
> > 
> > Can you check it has no regression wrt your suggested fix ?
> > 
> > Regards,
> > cvaroqui
> 
> Looks good, except the call to extract_hwe_from_path in setup_multipath
> must be before the call to select_pgfailback.

Ed,

So it sounds like I need Red Hat to issue me a hotfix for device-mapper-
multipath that includes the upstream patch or your patch for get
Clariion fallback working properly.

I'll ask Red Hat about this in my existing support case.

/Brian/

-- 
       Brian Long                      |         |           |
       IT Data Center Systems          |       .|||.       .|||.
       Cisco Linux Developer           |   ..:|||||||:...:|||||||:..
       Phone: (919) 392-7363           |   C i s c o   S y s t e m s

Re: Clariion CX600 automatic failback support

[Bernd Zeimetz]

Hi,

> So it sounds like I need Red Hat to issue me a hotfix for device-mapper-
> multipath that includes the upstream patch or your patch for get
> Clariion fallback working properly.

tested on a debian sarge machine and a CX400 today, actual multipath-tools 
from git, including the patch from 
https://www.redhat.com/archives/dm-devel/2005-November/msg00028.html

Seems to work very well, especially fine that the LUNs get trespassed to their 
default owner.

Many thanks to all who made this working.


Bernd

Re: Clariion CX600 automatic failback support

[Bernd Zeimetz]

Hi,

> Looks good, except the call to extract_hwe_from_path in setup_multipath
> must be before the call to select_pgfailback.
>

We have a ton of these log messages now.
From the last about 2 hours:

580x multipathd: 3600601f4a20c000099dc229956b0d711: switch to path group #1
714x multipathd: 3600601f4a20c000096dc229956b0d711: switch to path group #1
714x multipathd: 3600601f2a20c0000fb13caa656b0d711: switch to path group #1
  45x kernel: device-mapper: dm-emc: emc_pg_init: sending switch-over command

It happens only for these 3 of 8 luns.

Kernel is 2.6.13.4, devmapper 1.01, latest multipath-tools from git, evms 
2.5.3

zeus-1:/# cat /etc/multipath.conf
defaults {
        failback                                immediate
}


Since we had a lot of trouble with our CX400 (Hardware problems, but they're 
supposed to be fixed) I'm wondering what these messages mean, and if we still 
have hardware/software bugs somewhere.

Any ideas?
Thanks for your help!


Bernd

Re: Clariion CX600 automatic failback support

[Christophe Varoqui]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset=unknown-8bit, Size: 1298 bytes --]

On Fri, Nov 04, 2005 at 09:19:06AM +0100, Bernd Zeimetz wrote:
> Hi,
> 
> > Looks good, except the call to extract_hwe_from_path in setup_multipath
> > must be before the call to select_pgfailback.
> >
> 
> We have a ton of these log messages now.
> >From the last about 2 hours:
> 
> 580x multipathd: 3600601f4a20c000099dc229956b0d711: switch to path group #1
> 714x multipathd: 3600601f4a20c000096dc229956b0d711: switch to path group #1
> 714x multipathd: 3600601f2a20c0000fb13caa656b0d711: switch to path group #1
>   45x kernel: device-mapper: dm-emc: emc_pg_init: sending switch-over command
> 
> It happens only for these 3 of 8 luns.
> 
> Kernel is 2.6.13.4, devmapper 1.01, latest multipath-tools from git, evms 
> 2.5.3
> 
> zeus-1:/# cat /etc/multipath.conf
> defaults {
>         failback                                immediate
> }
> 
> 
> Since we had a lot of trouble with our CX400 (Hardware problems, but they're 
> supposed to be fixed) I'm wondering what these messages mean, and if we still 
> have hardware/software bugs somewhere.
> 
Yes, the test to determine if a switch is needed is sub-optimal :
The no activated PG case (all Enabled for exampleÃ) keeps triggering a switch.
I checked-in a patch to solve that issue.

Regards,
cvaroqui

RE: Clariion CX600 automatic failback support

[goggin, edward]

On Date: Fri, 4 Nov 2005 10:47:51 +0100
Christophe Varoqui <christophe.varoqui@free.fr> wrote

> On Fri, Nov 04, 2005 at 09:19:06AM +0100, Bernd Zeimetz wrote:
> > Hi,
> > 
> > > Looks good, except the call to extract_hwe_from_path in 
> setup_multipath
> > > must be before the call to select_pgfailback.
> > >
> > 
> > We have a ton of these log messages now.
> > >From the last about 2 hours:
> > 
> > 580x multipathd: 3600601f4a20c000099dc229956b0d711: switch 
> to path group #1
> > 714x multipathd: 3600601f4a20c000096dc229956b0d711: switch 
> to path group #1
> > 714x multipathd: 3600601f2a20c0000fb13caa656b0d711: switch 
> to path group #1
> >   45x kernel: device-mapper: dm-emc: emc_pg_init: sending 
> switch-over command
> > 
> > It happens only for these 3 of 8 luns.
> > 
> > Kernel is 2.6.13.4, devmapper 1.01, latest multipath-tools 
> from git, evms 
> > 2.5.3
> > 
> > zeus-1:/# cat /etc/multipath.conf
> > defaults {
> >         failback                                immediate
> > }
> > 
> > 
> > Since we had a lot of trouble with our CX400 (Hardware 
> problems, but they're 
> > supposed to be fixed) I'm wondering what these messages 
> mean, and if we still 
> > have hardware/software bugs somewhere.
> > 
> Yes, the test to determine if a switch is needed is sub-optimal :
> The no activated PG case (all Enabled for exampleÃ) keeps 
> triggering a switch.
> I checked-in a patch to solve that issue.

I've dealt with this problem by patching the device mapper
multipath driver in the kernel.

The patch below patches drivers/md/dm-mpath.c:multipath_status()
to return an active state for a path group which is either the
current path group or setup to be the next path group instead
of just returning an active state for a path group which is
currently active.

*** ../base/linux-2.6.14-rc4/drivers/md/dm-mpath.c	Mon Oct 10 20:19:19
2005
--- drivers/md/dm-mpath.c	Thu Nov  3 04:17:48 2005
***************
*** 1158,1164 ****
  		list_for_each_entry(pg, &m->priority_groups, list) {
  			if (pg->bypassed)
  				state = 'D';	/* Disabled */
! 			else if (pg == m->current_pg)
  				state = 'A';	/* Currently Active */
  			else
  				state = 'E';	/* Enabled */
--- 1158,1164 ----
  		list_for_each_entry(pg, &m->priority_groups, list) {
  			if (pg->bypassed)
  				state = 'D';	/* Disabled */
! 			else if ((pg == m->current_pg) || (pg ==
m->next_pg))
  				state = 'A';	/* Currently Active */
  			else
  				state = 'E';	/* Enabled */

RE: Clariion CX600 automatic failback support

[Christophe Varoqui]

> > Yes, the test to determine if a switch is needed is sub-optimal :
> > The no activated PG case (all Enabled for exampleÃ) keeps 
> > triggering a switch.
> > I checked-in a patch to solve that issue.
> 
> I've dealt with this problem by patching the device mapper
> multipath driver in the kernel.
> 
> The patch below patches drivers/md/dm-mpath.c:multipath_status()
> to return an active state for a path group which is either the
> current path group or setup to be the next path group instead
> of just returning an active state for a path group which is
> currently active.
> 
Right, it deserved to be done.

I'm inclined to keep the tools patch anyway : it wastes an "int" per
multipath but is brain-friendlier (as it avoids using mpp->nextpg to
represent the current setting *and* the best-rated setting).

Are interested parties ok with that ?

Regards,
cvaroqui

Re: Clariion CX600 automatic failback support

[Bernd Zeimetz]

Hi,

> I'm inclined to keep the tools patch anyway : it wastes an "int" per
> multipath but is brain-friendlier (as it avoids using mpp->nextpg to
> represent the current setting *and* the best-rated setting).
>
> Are interested parties ok with that ?

I'm ok with anything you prefer, as long as it works and it doesn't change 
again with the next kernel/tools-release ;)

Thank you,


Bernd

Re: Clariion CX600 automatic failback support

[Christophe Varoqui]

On sam, 2005-11-05 at 01:03 +0100, Bernd Zeimetz wrote:
> Hi,
> 
> > I'm inclined to keep the tools patch anyway : it wastes an "int" per
> > multipath but is brain-friendlier (as it avoids using mpp->nextpg to
> > represent the current setting *and* the best-rated setting).
> >
> > Are interested parties ok with that ?
> 
> I'm ok with anything you prefer, as long as it works and it doesn't change 
> again with the next kernel/tools-release ;)
> 
Want to write a regression test ?

Regards,
cvaroqui

Re: Clariion CX600 automatic failback support

[Bernd Zeimetz]

Hi,

> Want to write a regression test ?

Doubt that's really neccessary - at least not for us (which means I'll have to 
work on the other waiting projects first before doing anything else here). 
Due to the lack of the company where we've bought the machine (unfortunately 
not directly from EMC) in supporting us with their own 
closed-source-multipath-modules for an actual kernel.org kernel it took me 
more time than planned to get the CX running stable again. Many thanks again 
to Christophe, Edward and Kevin for the support & patches!

Within the next week I'll update your wiki with a completion of the 
DebianInstall page, which will include a link to an updated multipath-tools 
package (latest git) for Debian Sarge.

A few minutes ago I've added the CX to 
http://christophe.varoqui.free.fr/wiki/wakka.php?wiki=TestedEnvironments - I 
hope it'll be useful. Please let me know if I've forgotten anything.

Thanks again,

Bernd

RE: Clariion CX600 automatic failback support

[goggin, edward]

On Date: Fri, 04 Nov 2005 21:55:20 +0100
Christophe Varoqui <christophe.varoqui@free.fr> wrote

> Subject: RE: [dm-devel] Clariion CX600 automatic failback support
> To: device-mapper development <dm-devel@redhat.com>
> Message-ID: <1131137721.7463.12.camel@zezette>
> Content-Type: text/plain; charset=utf-8
> 
> 
> > > Yes, the test to determine if a switch is needed is sub-optimal :
> > > The no activated PG case (all Enabled for exampleÃ) keeps 
> > > triggering a switch.
> > > I checked-in a patch to solve that issue.
> > 
> > I've dealt with this problem by patching the device mapper
> > multipath driver in the kernel.
> > 
> > The patch below patches drivers/md/dm-mpath.c:multipath_status()
> > to return an active state for a path group which is either the
> > current path group or setup to be the next path group instead
> > of just returning an active state for a path group which is
> > currently active.
> > 
> Right, it deserved to be done.
> 
> I'm inclined to keep the tools patch anyway : it wastes an "int" per
> multipath but is brain-friendlier (as it avoids using mpp->nextpg to
> represent the current setting *and* the best-rated setting).
> 
> Are interested parties ok with that ?

Good point.  This slightly modified patch fixes the "*and*" case by only
returning active if either the particular path group is the current active
or there is no current active but the particular path group is set to be
the next (first) deferred active path group.

*** ../base/linux-2.6.14-rc4/drivers/md/dm-mpath.c	Mon Oct 10 20:19:19
2005
--- drivers/md/dm-mpath.c	Sat Nov  5 03:02:50 2005
***************
*** 1158,1164 ****
  		list_for_each_entry(pg, &m->priority_groups, list) {
  			if (pg->bypassed)
  				state = 'D';	/* Disabled */
! 			else if (pg == m->current_pg)
  				state = 'A';	/* Currently Active */
  			else
  				state = 'E';	/* Enabled */
--- 1158,1164 ----
  		list_for_each_entry(pg, &m->priority_groups, list) {
  			if (pg->bypassed)
  				state = 'D';	/* Disabled */
! 			else if ((pg == m->current_pg) || ((!m->current_pg)
&&(pg == m->next_pg)))
  				state = 'A';	/* Currently Active */
  			else
  				state = 'E';	/* Enabled */