multipath and StorageWorks HSG80

multipath and StorageWorks HSG80

[Steve Feehan]

Hello. I'm trying to figure out multipath on a StorageWorks HSG80
contoller. I'm using Ubuntu "Dapper" with the following software
versions:

  kernel:           2.6.15-20-server
  multipath-tools:  0.4.7-1ubuntu4

From what I've been reading[1], I gather that I need to use the
dm-hp-sw kernel module. So I applied the patches[2,3,4,5] to
the kernel and setup the /etc/multipath.conf as follows:

devices {

        device {
                vendor                  "DEC"
                product                 "HSG80"
                path_grouping_policy    group_by_serial
                path_checker            hp_sw
                features                "1 queue_if_no_path"
                hardware_handler        "1 hp_sw"
        }
}

multipaths {
        multipath {
                wwid    60001fe10015bf500009947159810015
                alias   red
        }
}

From what I can tell, it is basically working:

sfeehan@dogwood:~$ sudo multipath -ll
red (360001fe10015bf500009947159810015)
[size=203 GB][features=1 queue_if_no_path][hwhandler=1 hp_sw]
\_ round-robin 0 [prio=2][active]
 \_ 0:0:0:1 sda 8:0   [active][ready]
 \_ 0:0:1:1 sdb 8:16  [active][ready]
\_ round-robin 0 [prio=2][enabled]
 \_ 0:0:2:1 sdc 8:32  [active][ghost]
 \_ 0:0:3:1 sdd 8:48  [active][ghost]


And now the problem: I see quite a few errors in dmesg:


[42949392.230000] sd 0:0:2:1: Device not ready.
[42949392.230000] end_request: I/O error, dev sdc, sector 426583424
[42949392.230000] printk: 515 messages suppressed.
[42949392.230000] Buffer I/O error on device sdc, logical block
213291712
[42949392.230000] sd 0:0:2:1: Device not ready.
[42949392.230000] end_request: I/O error, dev sdc, sector 426583424
[42949392.230000] sd 0:0:3:1: Device not ready.
[42949392.230000] end_request: I/O error, dev sdd, sector 426583424
[42949392.230000] sd 0:0:3:1: Device not ready.
[42949392.230000] end_request: I/O error, dev sdd, sector 426583424


And when I put LVM on top of the multipath device I get lots
of errors:

sfeehan@dogwood:~$ sudo pvcreate /dev/mapper/red
  Physical volume "/dev/mapper/red" successfully created
sfeehan@dogwood:~$ sudo vgcreate vg0 /dev/mapper/red
  Found duplicate PV qDhK4gPTGtzXAm7jPQf0uf07a0FQs2qL: using /dev/sdb1
not /dev/sda1
  /dev/sdc: read failed after 0 of 2048 at 0: Input/output error
  /dev/sdd: read failed after 0 of 2048 at 0: Input/output error
  Found duplicate PV qDhK4gPTGtzXAm7jPQf0uf07a0FQs2qL: using /dev/sda1
not /dev/sdb1
  Found duplicate PV qDhK4gPTGtzXAm7jPQf0uf07a0FQs2qL: using /dev/sdb1
not /dev/sda1
  /dev/sdc: read failed after 0 of 1024 at 0: Input/output error
  /dev/sdc: read failed after 0 of 2048 at 0: Input/output error
  /dev/sdd: read failed after 0 of 1024 at 0: Input/output error
  /dev/sdd: read failed after 0 of 2048 at 0: Input/output error
  Volume group "vg0" successfully created
sfeehan@dogwood:~$
sfeehan@dogwood:~$
sfeehan@dogwood:~$ sudo lvcreate -L 10G -n test vg0
  Found duplicate PV qDhK4gPTGtzXAm7jPQf0uf07a0FQs2qL: using /dev/sdb1
not /dev/sda1
  /dev/sdc: read failed after 0 of 2048 at 0: Input/output error
  /dev/sdd: read failed after 0 of 2048 at 0: Input/output error
  Logical volume "test" created
sfeehan@dogwood:~$


So... what are the errors about? Are they significant? What
if anything can I do to resolve them?

Thanks.

Steve


[1] [dm-devel] StorageWorks multipath support
    http://www.redhat.com/archives/dm-devel/2005-June/msg00105.html

[2] [dm-devel] [PATCH RFT/RFC 1/4] export block layer helpers
    http://www.redhat.com/archives/dm-devel/2005-September/msg00097.html

[3] [dm-devel] [PATCH RFT/RFC 2/4] add dm_scsi helpers
    http://www.redhat.com/archives/dm-devel/2005-September/msg00098.html

[4] [dm-devel] [PATCH RFT/RFC 3/4] convert dm emc
    http://www.redhat.com/archives/dm-devel/2005-September/msg00099.html

[5] [dm-devel] [PATCH RFT/RFC 4/4] add dm-hp-sw hw handler
    http://www.redhat.com/archives/dm-devel/2005-September/msg00100.html

-- 
Steve Feehan

Re: multipath and StorageWorks HSG80

[Bernd Zeimetz]

Steve Feehan wrote:
> Hello. I'm trying to figure out multipath on a StorageWorks HSG80
> contoller. I'm using Ubuntu "Dapper" with the following software
>   
there was a thread about the HSG80s few days ago.
http://www.redhat.com/archives/dm-devel/2006-April/msg00025.html
probably that answers all your questions.


Actually I'd really like to have them supported well, chances are good
that we'll get some of them soon :\


Best regards,


Bernd

Re: multipath and StorageWorks HSG80

[Christophe Varoqui]

Steve Feehan a écrit :
> Hello. I'm trying to figure out multipath on a StorageWorks HSG80
> contoller. I'm using Ubuntu "Dapper" with the following software
> versions:
>
>   kernel:           2.6.15-20-server
>   multipath-tools:  0.4.7-1ubuntu4
>
> >From what I've been reading[1], I gather that I need to use the
> dm-hp-sw kernel module. So I applied the patches[2,3,4,5] to
> the kernel and setup the /etc/multipath.conf as follows:
>
> devices {
>
>         device {
>                 vendor                  "DEC"
>                 product                 "HSG80"
>                 path_grouping_policy    group_by_serial
>                 path_checker            hp_sw
>                 features                "1 queue_if_no_path"
>                 hardware_handler        "1 hp_sw"
>         }
> }
>
> multipaths {
>         multipath {
>                 wwid    60001fe10015bf500009947159810015
>                 alias   red
>         }
> }
>
> >From what I can tell, it is basically working:
>
> sfeehan@dogwood:~$ sudo multipath -ll
> red (360001fe10015bf500009947159810015)
> [size=203 GB][features=1 queue_if_no_path][hwhandler=1 hp_sw]
> \_ round-robin 0 [prio=2][active]
>  \_ 0:0:0:1 sda 8:0   [active][ready]
>  \_ 0:0:1:1 sdb 8:16  [active][ready]
> \_ round-robin 0 [prio=2][enabled]
>  \_ 0:0:2:1 sdc 8:32  [active][ghost]
>  \_ 0:0:3:1 sdd 8:48  [active][ghost]
>
>
> And now the problem: I see quite a few errors in dmesg:
>
>
> [42949392.230000] sd 0:0:2:1: Device not ready.
> [42949392.230000] end_request: I/O error, dev sdc, sector 426583424
> [42949392.230000] printk: 515 messages suppressed.
> [42949392.230000] Buffer I/O error on device sdc, logical block
> 213291712
> [42949392.230000] sd 0:0:2:1: Device not ready.
> [42949392.230000] end_request: I/O error, dev sdc, sector 426583424
> [42949392.230000] sd 0:0:3:1: Device not ready.
> [42949392.230000] end_request: I/O error, dev sdd, sector 426583424
> [42949392.230000] sd 0:0:3:1: Device not ready.
> [42949392.230000] end_request: I/O error, dev sdd, sector 426583424
>
>
> And when I put LVM on top of the multipath device I get lots
> of errors:
>
> sfeehan@dogwood:~$ sudo pvcreate /dev/mapper/red
>   Physical volume "/dev/mapper/red" successfully created
> sfeehan@dogwood:~$ sudo vgcreate vg0 /dev/mapper/red
>   Found duplicate PV qDhK4gPTGtzXAm7jPQf0uf07a0FQs2qL: using /dev/sdb1
> not /dev/sda1
>   /dev/sdc: read failed after 0 of 2048 at 0: Input/output error
>   /dev/sdd: read failed after 0 of 2048 at 0: Input/output error
>   Found duplicate PV qDhK4gPTGtzXAm7jPQf0uf07a0FQs2qL: using /dev/sda1
> not /dev/sdb1
>   Found duplicate PV qDhK4gPTGtzXAm7jPQf0uf07a0FQs2qL: using /dev/sdb1
> not /dev/sda1
>   /dev/sdc: read failed after 0 of 1024 at 0: Input/output error
>   /dev/sdc: read failed after 0 of 2048 at 0: Input/output error
>   /dev/sdd: read failed after 0 of 1024 at 0: Input/output error
>   /dev/sdd: read failed after 0 of 2048 at 0: Input/output error
>   Volume group "vg0" successfully created
> sfeehan@dogwood:~$
> sfeehan@dogwood:~$
> sfeehan@dogwood:~$ sudo lvcreate -L 10G -n test vg0
>   Found duplicate PV qDhK4gPTGtzXAm7jPQf0uf07a0FQs2qL: using /dev/sdb1
> not /dev/sda1
>   /dev/sdc: read failed after 0 of 2048 at 0: Input/output error
>   /dev/sdd: read failed after 0 of 2048 at 0: Input/output error
>   Logical volume "test" created
> sfeehan@dogwood:~$
>
>
> So... what are the errors about? Are they significant? What
> if anything can I do to resolve them?
>
> Thanks.
>
> Steve
>
>
> [1] [dm-devel] StorageWorks multipath support
>     http://www.redhat.com/archives/dm-devel/2005-June/msg00105.html
>
> [2] [dm-devel] [PATCH RFT/RFC 1/4] export block layer helpers
>     http://www.redhat.com/archives/dm-devel/2005-September/msg00097.html
>
> [3] [dm-devel] [PATCH RFT/RFC 2/4] add dm_scsi helpers
>     http://www.redhat.com/archives/dm-devel/2005-September/msg00098.html
>
> [4] [dm-devel] [PATCH RFT/RFC 3/4] convert dm emc
>     http://www.redhat.com/archives/dm-devel/2005-September/msg00099.html
>
> [5] [dm-devel] [PATCH RFT/RFC 4/4] add dm-hp-sw hw handler
>     http://www.redhat.com/archives/dm-devel/2005-September/msg00100.html
>
>   
Overall, that's very good howto material, if you care to feed the wiki.

The IO errors are due to lvm tools trying to scan the individual paths, 
including those not usable, where they should only care about the 
multipath devices.

You should blacklist the individual path in the lvm config.


Regards,
cvaroqui

Re: multipath and StorageWorks HSG80

[Steve Feehan]

On Thu, Apr 20, 2006 at 10:07:50PM +0200, Christophe Varoqui wrote:

> Overall, that's very good howto material, if you care to feed the wiki.

OK, I will see what I can do.

> The IO errors are due to lvm tools trying to scan the individual paths, 
> including those not usable, where they should only care about the 
> multipath devices.
>
> You should blacklist the individual path in the lvm config.

Yes, I put the following in /etc/lvm/lvm.conf:

filter = [ "r|/dev/cdrom|", "r|/dev/sd|" ]

And that makes the errors go away for the LVM case. But I still 
get the kernel errors during bootup.

It's too soon to say for sure, but so far I don't seem to be
getting any errors once the system has booted. Perhaps it's just
during the early stages of booting and not an issue after that?

If anyone would like to see the full dmesg output for a boot
I'd be glad to send it.

And I haven't even begun to test actually failing over the
controllers. I imagine I'll have more problems/questions shortly. :)

Thanks.

-- 
Steve Feehan

Re: multipath and StorageWorks HSG80

[Steve Feehan]

On Thu, Apr 20, 2006 at 10:02:07PM +0200, Bernd Zeimetz wrote:
> Steve Feehan wrote:
> > Hello. I'm trying to figure out multipath on a StorageWorks HSG80
> > contoller. I'm using Ubuntu "Dapper" with the following software
> >   
> there was a thread about the HSG80s few days ago.
> http://www.redhat.com/archives/dm-devel/2006-April/msg00025.html
> probably that answers all your questions.

Yes, I saw that. The particular problem that I saw there
was the misreporting of unit size. The Ubuntu init script
takes care of this and all of my units report the same 
size.

> Actually I'd really like to have them supported well, chances are good
> that we'll get some of them soon :\
> 
> 
> Best regards,
> 
> 
> Bernd
> 
> --
> dm-devel mailing list
> dm-devel@redhat.com
> https://www.redhat.com/mailman/listinfo/dm-devel

-- 
Steve Feehan

Re: multipath and StorageWorks HSG80

[Christophe Varoqui]

Steve Feehan a écrit :
> On Thu, Apr 20, 2006 at 10:02:07PM +0200, Bernd Zeimetz wrote:
>   
>> Steve Feehan wrote:
>>     
>>> Hello. I'm trying to figure out multipath on a StorageWorks HSG80
>>> contoller. I'm using Ubuntu "Dapper" with the following software
>>>   
>>>       
>> there was a thread about the HSG80s few days ago.
>> http://www.redhat.com/archives/dm-devel/2006-April/msg00025.html
>> probably that answers all your questions.
>>     
>
> Yes, I saw that. The particular problem that I saw there
> was the misreporting of unit size. The Ubuntu init script
> takes care of this and all of my units report the same 
> size.
>
>   
That's certainly where the IO errors are triggered at bootup.
Can you send the code snippet ?

Regards,
cvaroqui

Re: multipath and StorageWorks HSG80

[Steve Feehan]

On Thu, Apr 20, 2006 at 10:53:00PM +0200, Christophe Varoqui wrote:
> Steve Feehan a ?crit :
> >On Thu, Apr 20, 2006 at 10:02:07PM +0200, Bernd Zeimetz wrote:
> >  
> >>Steve Feehan wrote:
> >>    
> >>>Hello. I'm trying to figure out multipath on a StorageWorks HSG80
> >>>contoller. I'm using Ubuntu "Dapper" with the following software
> >>>  
> >>>      
> >>there was a thread about the HSG80s few days ago.
> >>http://www.redhat.com/archives/dm-devel/2006-April/msg00025.html
> >>probably that answers all your questions.
> >>    
> >
> >Yes, I saw that. The particular problem that I saw there
> >was the misreporting of unit size. The Ubuntu init script
> >takes care of this and all of my units report the same 
> >size.
> >
> >  
> That's certainly where the IO errors are triggered at bootup.
> Can you send the code snippet ?

You mean from the init script? Here it is:

hsg80_init() {
        dummy_capa=2097152

        for i in $(grep -rl 2097152 /sys/block/sd*/size|awk -F/ '{print
$4}')
        do
                echo
                echo "##################################"
                echo "FIXING HSG80 device: $i"
                echo
                sg_start -start /dev/$i
                sleep 1
                echo 1>/sys/block/$i/device/rescan
        done
}

I put the echo's in so that I could see if and for which devices
it was being called. And yes, it's called for the two "ghost"
devices (sdc and sdd).


But if you look at boot output, the errors begin well before
the multipath-tools init script runs. 

Here is a trimmed excerpt:


[42949387.600000] qla2200 0000:03:01.0: Found an ISP2200, irq 169,
iobase 0xf8826000
[42949387.700000] qla2200 0000:03:01.0: Configuring PCI space...
[42949387.780000] qla2200 0000:03:01.0: Configure NVRAM parameters...
[42949387.940000] qla2200 0000:03:01.0: Verifying loaded RISC code...
[42949388.200000] qla2200 0000:03:01.0: LIP reset occured (0).
[42949388.270000] qla2200 0000:03:01.0: Waiting for LIP to complete...
[42949390.910000] qla2200 0000:03:01.0: LOOP UP detected (1 Gbps).
[42949390.990000] qla2200 0000:03:01.0: Topology - (F_Port), Host Loop
address 0xffff
[42949391.170000] scsi0 : qla2xxx
[42949391.210000] qla2200 0000:03:01.0:
[42949391.210000]  QLogic Fibre Channel HBA Driver: 8.01.03-k
[42949391.210000]   QLogic QLA22xx -
[42949391.210000]   ISP2200: PCI (66 MHz) @ 0000:03:01.0 hdma-, host#=0,
fw=2.02.08 TP
[42949391.480000] scsi: unknown device type 12
[42949391.540000]   Vendor: DEC       Model: HSG80CCL          Rev: V87F
[42949391.620000]   Type:   RAID                               ANSI SCSI
revision: 02
[42949391.730000]   Vendor: DEC       Model: HSG80             Rev: V87F
[42949391.810000]   Type:   Direct-Access                      ANSI SCSI
revision: 02
[42949391.910000] usbcore: registered new driver usbfs
[42949391.910000] scsi: unknown device type 12
[42949391.910000]   Vendor: DEC       Model: HSG80CCL          Rev: V87F
[42949391.910000]   Type:   RAID                               ANSI SCSI
revision: 02
[42949391.920000]   Vendor: DEC       Model: HSG80             Rev: V87F
[42949391.920000]   Type:   Direct-Access                      ANSI SCSI
revision: 02
[42949392.400000] usbcore: registered new driver hub
[42949392.470000] scsi: unknown device type 12
[42949392.520000]   Vendor: DEC       Model: HSG80CCL          Rev: V87F
[42949392.610000]   Type:   RAID                               ANSI SCSI
revision: 02
[42949392.710000] ACPI: PCI Interrupt 0000:01:00.0[D] -> GSI 19 (level,
low) -> IRQ 177
[42949394.820000] SCSI device sda: 426583554 512-byte hdwr sectors
(218411 MB)
[42949394.910000] SCSI device sda: drive cache: write through
[42949394.980000]  sda: unknown partition table
[42949395.050000] sd 0:0:0:1: Attached scsi disk sda
[42949395.120000] SCSI device sdb: 426583554 512-byte hdwr sectors
(218411 MB)
[42949395.210000] SCSI device sdb: drive cache: write through
[42949395.290000] SCSI device sdb: 426583554 512-byte hdwr sectors
(218411 MB)
[42949395.380000] SCSI device sdb: drive cache: write through
[42949395.450000]  sdb: unknown partition table
[42949395.510000] sd 0:0:1:1: Attached scsi disk sdb
[42949395.580000] sdc : READ CAPACITY failed.
[42949395.580000] sdc : status=1, message=00, host=0, driver=08
[42949395.710000] sd: Current: sense key: Not Ready
[42949395.770000]     Additional sense: Logical unit not ready,
initializing cmd. required
[42949395.870000] sdc: asking for cache data failed
[42949395.930000] sdc: assuming drive cache: write through
[42949396.000000] sdc : READ CAPACITY failed.
[42949396.000000] sdc : status=1, message=00, host=0, driver=08
[42949396.130000] sd: Current: sense key: Not Ready
[42949396.190000]     Additional sense: Logical unit not ready,
initializing cmd. required
[42949396.300000] sdc: asking for cache data failed
[42949396.360000] sdc: assuming drive cache: write through
[42949396.430000]  sdc:<6>sd 0:0:2:1: Device not ready.
[42949396.490000] end_request: I/O error, dev sdc, sector 0
[42949396.560000] Buffer I/O error on device sdc, logical block 0
[42949396.640000] sd 0:0:2:1: Device not ready.
[42949396.700000] end_request: I/O error, dev sdc, sector 0
[42949396.770000] Buffer I/O error on device sdc, logical block 0
[42949396.840000]  unable to read partition table
[42949396.900000] sd 0:0:2:1: Attached scsi disk sdc
[42949396.970000] sdd : READ CAPACITY failed.
[42949396.970000] sdd : status=1, message=00, host=0, driver=08
[42949397.100000] sd 0:0:2:1: Device not ready.
[42949397.100000] end_request: I/O error, dev sdc, sector 2097024
[42949397.100000] Buffer I/O error on device sdc, logical block 262128
[42949397.310000] sd 0:0:2:1: Device not ready.


.... and then hundreds of lines of such errors ....

Begin: Running /scripts/local-bottom ...
Done.
Done.
Begin: Running /scripts/init-bottom ...
Done.
 * version 2.86 booting
Starting multipath
##################################
FIXING HSG80 device: sdc

[42949413.900000] SCSI device sdc: 426583554 512-byte hdwr sectors
(218411 MB)
[42949413.990000] SCSI device sdc: drive cache: write through

##################################
FIXING HSG80 device: sdd

[42949415.090000] SCSI device sdd: 426583554 512-byte hdwr sectors
(218411 MB)
[42949415.180000] SCSI device sdd: drive cache: write through
[42949415.440000] device-mapper: hp_sw version 0.4 loaded
[42949415.520000] device-mapper: dm-round-robin version 1.0.0 loaded
 done.
 * Starting basic networking...
 * [ ok ]
 * Starting kernel event manager...
 * [ ok ]
 * Loading hardware drivers... [42949416.810000] device-mapper: hp_sw:
 * queueing START_STOP command on 8:16
[42949417.030000]  0:0:0:0: Attached scsi generic sg0 type 12
[42949417.100000] sd 0:0:0:1: Attached scsi generic sg1 type 0
[42949417.170000]  0:0:1:0: Attached scsi generic sg2 type 12
[42949417.240000] sd 0:0:1:1: Attached scsi generic sg3 type 0
[42949417.320000] sd 0:0:3:1: Device not ready.
[42949417.320000] end_request: I/O error, dev sdd, sector 426583424
[42949417.320000] printk: 40 messages suppressed.


.... and the errors continue for the rest of the boot process.
Once the file system is mounted, no more errors. At least not
for the short period that the system has been up.

-- 
Steve Feehan

Re: multipath and StorageWorks HSG80

[Eddie Williams]

On Thu, 2006-04-20 at 16:44 -0400, Steve Feehan wrote:
> On Thu, Apr 20, 2006 at 10:07:50PM +0200, Christophe Varoqui wrote:

> And that makes the errors go away for the LVM case. But I still 
> get the kernel errors during bootup.
> 

Did you rebuild your RAM disk?  If not it will scan LVM using the conf
file from the last build.

Re: multipath and StorageWorks HSG80

[Christophe Varoqui]

>> That's certainly where the IO errors are triggered at bootup.
>> Can you send the code snippet ?
>>     
>
> You mean from the init script? Here it is:
>
> hsg80_init() {
>         dummy_capa=2097152
>
>         for i in $(grep -rl 2097152 /sys/block/sd*/size|awk -F/ '{print
> $4}')
>         do
>                 echo
>                 echo "##################################"
>                 echo "FIXING HSG80 device: $i"
>                 echo
>                 sg_start -start /dev/$i
>                 sleep 1
>                 echo 1>/sys/block/$i/device/rescan
>         done
> }
>
> I put the echo's in so that I could see if and for which devices
> it was being called. And yes, it's called for the two "ghost"
> devices (sdc and sdd).
>
>
> But if you look at boot output, the errors begin well before
> the multipath-tools init script runs. 
>
>   
Right, this script can't be the culprit.
> Here is a trimmed excerpt:
>
>
> (218411 MB)
> [42949395.380000] SCSI device sdb: drive cache: write through
> [42949395.450000]  sdb: unknown partition table
> [42949395.510000] sd 0:0:1:1: Attached scsi disk sdb
> [42949395.580000] sdc : READ CAPACITY failed.
> [42949395.580000] sdc : status=1, message=00, host=0, driver=08
> [42949395.710000] sd: Current: sense key: Not Ready
> [42949395.770000]     Additional sense: Logical unit not ready,
> initializing cmd. required
> [42949395.870000] sdc: asking for cache data failed
> [42949395.930000] sdc: assuming drive cache: write through
> [42949396.000000] sdc : READ CAPACITY failed.
> [42949396.000000] sdc : status=1, message=00, host=0, driver=08
> [42949396.130000] sd: Current: sense key: Not Ready
> [42949396.190000]     Additional sense: Logical unit not ready,
> initializing cmd. required
> [42949396.300000] sdc: asking for cache data failed
> [42949396.360000] sdc: assuming drive cache: write through
> [42949396.430000]  sdc:<6>sd 0:0:2:1: Device not ready.
> [42949396.490000] end_request: I/O error, dev sdc, sector 0
>   
The scsi driver probing seems to fail with io errors on ghost paths.
If so, you can consider it harmless.

Regards,
cvaroqui

Re: multipath and StorageWorks HSG80

[Steve Feehan]

On Thu, Apr 20, 2006 at 05:10:44PM -0400, Eddie Williams wrote:
> On Thu, 2006-04-20 at 16:44 -0400, Steve Feehan wrote:
> > On Thu, Apr 20, 2006 at 10:07:50PM +0200, Christophe Varoqui wrote:
> 
> > And that makes the errors go away for the LVM case. But I still 
> > get the kernel errors during bootup.
> > 
> 
> Did you rebuild your RAM disk?  If not it will scan LVM using the conf
> file from the last build.

No, I hadn't considered that. But looking at the content of the
initramfs there is not a copy of lvm.conf in there.

But I have since rebuilt the initramfs and it did not make a
difference.

I have not customized the default Ubuntu initramfs (other than
building it for the patched kernel that I needed to build to get
the dm-hp-sw module). Should I be doing anything special in the
initramfs? Note that I'm not trying to put the / file system on
a multipath device... I'm not nearly that ambitious. :)

Also, I noticed this when the device-mapper modules are loaded
from the initramfs:


Begin: Loading e[42949384.440000] SCSI subsystem initialized
ssential drivers[42949384.500000] device-mapper: 4.4.0-ioctl
(2005-01-12) initialised: dm-devel@redhat.com
... ...
[42949384.620000] device-mapper: dm-multipath version 1.0.4 loaded
[42949384.710000] dm_hp_sw: Unknown parameter `dm-hp-sw.c'
Done.


But looking at the boot text, there was something wrong with the
loading of dm-hp-sw from the initramfs. It appears that the
dm-hp-sw module was not successfully loaded until much later in
the boot sequence.

I tracked this down and fixed the problem. But still, even with
dm-hp-sw loading before the SCSI devices are probed, I get a ton
of SCSI errors trying to access the "ghost" devices.

-- 
Steve Feehan

Re: multipath and StorageWorks HSG80

[Philip R. Auld]

Rumor has it that on Fri, Apr 21, 2006 at 09:40:55AM -0400 Steve Feehan said:
> 
> I tracked this down and fixed the problem. But still, even with
> dm-hp-sw loading before the SCSI devices are probed, I get a ton
> of SCSI errors trying to access the "ghost" devices.

Those should be expected. The SCSI layers are trying to configure the
devices (read capacity, partition tables etc). The dm-hp-sw module
is not in that path so it can't do anything about that. They should be
harmless. 
 
This is an architectural artifact of doing the multipathing above the
SCSI layer.


Cheers,

Phil

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

-- 
Philip R. Auld, Ph.D.  	        	       Egenera, Inc.    
Software Architect                            165 Forest St.
(508) 858-2628                            Marlboro, MA 01752

Re: multipath and StorageWorks HSG80

[Steve Feehan]

On Fri, Apr 21, 2006 at 10:09:42AM -0400, Philip R. Auld wrote:
> Rumor has it that on Fri, Apr 21, 2006 at 09:40:55AM -0400 Steve Feehan said:
> > 
> > I tracked this down and fixed the problem. But still, even with
> > dm-hp-sw loading before the SCSI devices are probed, I get a ton
> > of SCSI errors trying to access the "ghost" devices.
> 
> Those should be expected. The SCSI layers are trying to configure the
> devices (read capacity, partition tables etc). The dm-hp-sw module
> is not in that path so it can't do anything about that. They should be
> harmless. 
>  
> This is an architectural artifact of doing the multipathing above the
> SCSI layer.

OK. Then as Christophe Varoqui also suggested, I will just /try/
to not worry too much. ;)

Thanks.

-- 
Steve Feehan