|
Replies:
9
-
Last Post:
Jan 25, 2006 1:38 PM
by: Leon Koll
|
|
|
|
|
|
|
SVM question: what am I doing wrong?
Posted:
Jan 2, 2006 2:04 PM
|
|
[No Body]
|
|
|
Tony Nguyen
Truong.Q.Nguyen@Sun....
|
|
|
|
Re: SVM question: what am I doing wrong?
Posted:
Jan 2, 2006 2:04 PM
in response to: Guest
|
|
Hi Leon,
Sorry for a late response but it was the holiday break here in the US
and we just got back today. I did a quick search but couldn't find
anything related to what you're seeing. I have some comments and
questions inline to help us figure out this issue.
I'm also cc'ing the lvm-discuss alias in hope that other SVM engineers
can give us additional insight.
-tony
>Hi Tony,
>I found your e-mail in OpenSolaris LVM forum.
>I have a quextion related to Solaris 10 SVM - could you please help me.
>I didn't check this in Solaris Express and didn't send the bug report
>to Sun Support that's why I need your advice.
>The problem is: I cannot add a device to the metaset in S10/x86 01/06.
>Here is my scenario:
># metadb -a -f c0d0s3 c0d0s4 c0d0s5
># metadb
> flags first blk block count
> a m pc luo 16 8192 /dev/dsk/c0d0s3
> a pc luo 16 8192 /dev/dsk/c0d0s4
> a pc luo 16 8192 /dev/dsk/c0d0s5
># metaset -s test -a
>/dev/rdsk/c4t5849562D494E432058495631303330202020202031303139302020202020 2
>033d0
># metaset -s test
>
>Set name = test, Set number = 1
>
>Host Owner
> chuk Yes
>
>Drive
>Dbase
>
>/dev/dsk/c4t5849562D494E4320584956313033302020202020313031393020202020202 0
>33d0 Yes
>
># metaset -s test -a
>/dev/rdsk/c4t5849562D494E432058495631303330202020202031303139302020202020 2
>034d0
>metaset: chuk: /: No such file or directory
>
>both dd # metadb
> flags first blk block count
> a m pc luo 16 8192 /dev/dsk/c0d0s3
> a pc luo 16 8192 /dev/dsk/c0d0s4
> a pc luo 16 8192 /dev/dsk/c0d0s5
># metaset -s test -a
>/dev/rdsk/c4t5849562D494E432058495631303330202020202031303139302020202020 2
>033d0
># metaset -s test
>
>Set name = test, Set number = 1
>
>Host Owner
> chuk Yes
>
>Drive
>Dbase
>
>Both
>/dev/dsk/c4t5849562D494E4320584956313033302020202020313031393020202020202 0
>33d0 and /dev/rdsk/c4t5849562D494E4320584956313033302020202020313031393020202020202
>033d0 are 1.5TB mpxio disks with EFI label and slice 0 starting from
>block 34, occupyimng the whole disk.
>
>What am I doing wrong? Is it a bug? Or the metasets are not supported
>in non-SunCluster environment.
>
>
Metasets or sets are supported in non-SunCluster environment. In fact,
we recommend metasets to help ease management of metadevices, except
metadevices that are used for root mirroring.
Let me do a quick recap to make sure that I understand the problem here.
The machine is running stock S10 01/06 (S10 Update 1) with mpxio.
You can add one of the mpxio disks (the disk ended with 33d0) but not
the other one.
Is there any differences between these mpxio disks (e.g. disk size,
storage unit model)?
Can you add the regular drives (non-multipath drives) to the set?
>I had the similar problem with S10/x86 03/05 with SunCluster installed.
>Then I installed S10 01/06 and decided to check things before I
>install the Sun Cluster.
>
>
Can you provide the explorer output so I can see the system's
configuration info? The explorer's script is at
http://classic.sunsolve.sun.com/pub-cgi/show.pl?target=explorer/explorer
>Your opinion will be appreciated. I need to build NFS HA cluster based
>on multi-tera volumes using SVM so I need to know is there any
>showstopper.
>
>
>
Cool.
_______________________________________________
lvm-discuss mailing list
lvm-discuss at opensolaris dot org
|
|
|
|
|
Posts:
74
From:
Registered:
3/9/05
|
|
|
|
Re: Re: SVM question: what am I doing wrong?
Posted:
Jan 2, 2006 2:20 PM
in response to: Tony Nguyen
|
|
The other thing we can do is to check the device ids for both disks and
make sure they are *seen* by the system and format.
Steve
Tony Nguyen wrote:
> Hi Leon,
>
> Sorry for a late response but it was the holiday break here in the US
> and we just got back today. I did a quick search but couldn't find
> anything related to what you're seeing. I have some comments and
> questions inline to help us figure out this issue.
>
> I'm also cc'ing the lvm-discuss alias in hope that other SVM engineers
> can give us additional insight.
>
> -tony
>
>> Hi Tony,
>> I found your e-mail in OpenSolaris LVM forum.
>> I have a quextion related to Solaris 10 SVM - could you please help me.
>> I didn't check this in Solaris Express and didn't send the bug report
>> to Sun Support that's why I need your advice.
>> The problem is: I cannot add a device to the metaset in S10/x86 01/06.
>> Here is my scenario:
>> # metadb -a -f c0d0s3 c0d0s4 c0d0s5
>> # metadb
>> flags first blk block count
>> a m pc luo 16 8192 /dev/dsk/c0d0s3
>> a pc luo 16 8192 /dev/dsk/c0d0s4
>> a pc luo 16 8192 /dev/dsk/c0d0s5
>> # metaset -s test -a
>> /dev/rdsk/c4t5849562D494E4320584956313033302020202020313031393020202020202
>>
>> 033d0
>> # metaset -s test
>>
>> Set name = test, Set number = 1
>>
>> Host Owner
>> chuk Yes
>>
>> Drive
>> Dbase
>>
>> /dev/dsk/c4t5849562D494E43205849563130333020202020203130313930202020202020
>>
>> 33d0 Yes
>>
>> # metaset -s test -a
>> /dev/rdsk/c4t5849562D494E4320584956313033302020202020313031393020202020202
>>
>> 034d0
>> metaset: chuk: /: No such file or directory
>>
>> both dd # metadb
>> flags first blk block count
>> a m pc luo 16 8192 /dev/dsk/c0d0s3
>> a pc luo 16 8192 /dev/dsk/c0d0s4
>> a pc luo 16 8192 /dev/dsk/c0d0s5
>> # metaset -s test -a
>> /dev/rdsk/c4t5849562D494E4320584956313033302020202020313031393020202020202
>>
>> 033d0
>> # metaset -s test
>>
>> Set name = test, Set number = 1
>>
>> Host Owner
>> chuk Yes
>>
>> Drive
>> Dbase
>>
>> Both
>> /dev/dsk/c4t5849562D494E43205849563130333020202020203130313930202020202020
>>
>> 33d0 and
>> /dev/rdsk/c4t5849562D494E4320584956313033302020202020313031393020202020202
>>
>> 033d0 are 1.5TB mpxio disks with EFI label and slice 0 starting from
>> block 34, occupyimng the whole disk.
>>
>> What am I doing wrong? Is it a bug? Or the metasets are not supported
>> in non-SunCluster environment.
>>
>>
> Metasets or sets are supported in non-SunCluster environment. In fact,
> we recommend metasets to help ease management of metadevices, except
> metadevices that are used for root mirroring.
>
> Let me do a quick recap to make sure that I understand the problem here.
>
> The machine is running stock S10 01/06 (S10 Update 1) with mpxio.
> You can add one of the mpxio disks (the disk ended with 33d0) but not
> the other one.
>
> Is there any differences between these mpxio disks (e.g. disk size,
> storage unit model)?
> Can you add the regular drives (non-multipath drives) to the set?
>
>> I had the similar problem with S10/x86 03/05 with SunCluster installed.
>> Then I installed S10 01/06 and decided to check things before I
>> install the Sun Cluster.
>>
>>
>
> Can you provide the explorer output so I can see the system's
> configuration info? The explorer's script is at
>
> http://classic.sunsolve.sun.com/pub-cgi/show.pl?target=explorer/explorer
>
>> Your opinion will be appreciated. I need to build NFS HA cluster based
>> on multi-tera volumes using SVM so I need to know is there any
>> showstopper.
>>
>>
>>
> Cool.
>
>
> _______________________________________________
> lvm-discuss mailing list
> lvm-discuss at opensolaris dot org
_______________________________________________
lvm-discuss mailing list
lvm-discuss at opensolaris dot org
|
|
|
|
|
Posts:
480
From:
Registered:
3/9/05
|
|
|
|
Re: Re: SVM question: what am I doing wrong?
Posted:
Jan 2, 2006 3:52 PM
in response to: Tony Nguyen
|
|
The deviceid on the disk appears suspicious. The numbers 202020 are the
ascii. So I would expect a target address to be more like
c4t<port_number>__XIV1030___1019________3
What disks are these ? Is there a firmware upgrade ? Also could you try
with S10U1 since that is now available.
-Sanjay
Tony Nguyen wrote:
> Hi Leon,
>
> Sorry for a late response but it was the holiday break here in the US
> and we just got back today. I did a quick search but couldn't find
> anything related to what you're seeing. I have some comments and
> questions inline to help us figure out this issue.
>
> I'm also cc'ing the lvm-discuss alias in hope that other SVM engineers
> can give us additional insight.
>
> -tony
>
>> Hi Tony,
>> I found your e-mail in OpenSolaris LVM forum.
>> I have a quextion related to Solaris 10 SVM - could you please help me.
>> I didn't check this in Solaris Express and didn't send the bug report
>> to Sun Support that's why I need your advice.
>> The problem is: I cannot add a device to the metaset in S10/x86 01/06.
>> Here is my scenario:
>> # metadb -a -f c0d0s3 c0d0s4 c0d0s5
>> # metadb
>> flags first blk block count
>> a m pc luo 16 8192 /dev/dsk/c0d0s3
>> a pc luo 16 8192 /dev/dsk/c0d0s4
>> a pc luo 16 8192 /dev/dsk/c0d0s5
>> # metaset -s test -a
>> /dev/rdsk/c4t5849562D494E4320584956313033302020202020313031393020202020202
>>
>> 033d0
>> # metaset -s test
>>
>> Set name = test, Set number = 1
>>
>> Host Owner
>> chuk Yes
>>
>> Drive
>> Dbase
>>
>> /dev/dsk/c4t5849562D494E43205849563130333020202020203130313930202020202020
>>
>> 33d0 Yes
>>
>> # metaset -s test -a
>> /dev/rdsk/c4t5849562D494E4320584956313033302020202020313031393020202020202
>>
>> 034d0
>> metaset: chuk: /: No such file or directory
>>
>> both dd # metadb
>> flags first blk block count
>> a m pc luo 16 8192 /dev/dsk/c0d0s3
>> a pc luo 16 8192 /dev/dsk/c0d0s4
>> a pc luo 16 8192 /dev/dsk/c0d0s5
>> # metaset -s test -a
>> /dev/rdsk/c4t5849562D494E4320584956313033302020202020313031393020202020202
>>
>> 033d0
>> # metaset -s test
>>
>> Set name = test, Set number = 1
>>
>> Host Owner
>> chuk Yes
>>
>> Drive
>> Dbase
>>
>> Both
>> /dev/dsk/c4t5849562D494E43205849563130333020202020203130313930202020202020
>>
>> 33d0 and
>> /dev/rdsk/c4t5849562D494E4320584956313033302020202020313031393020202020202
>>
>> 033d0 are 1.5TB mpxio disks with EFI label and slice 0 starting from
>> block 34, occupyimng the whole disk.
>>
>> What am I doing wrong? Is it a bug? Or the metasets are not supported
>> in non-SunCluster environment.
>>
>>
> Metasets or sets are supported in non-SunCluster environment. In fact,
> we recommend metasets to help ease management of metadevices, except
> metadevices that are used for root mirroring.
>
> Let me do a quick recap to make sure that I understand the problem here.
>
> The machine is running stock S10 01/06 (S10 Update 1) with mpxio.
> You can add one of the mpxio disks (the disk ended with 33d0) but not
> the other one.
>
> Is there any differences between these mpxio disks (e.g. disk size,
> storage unit model)?
> Can you add the regular drives (non-multipath drives) to the set?
>
>> I had the similar problem with S10/x86 03/05 with SunCluster installed.
>> Then I installed S10 01/06 and decided to check things before I
>> install the Sun Cluster.
>>
>>
>
> Can you provide the explorer output so I can see the system's
> configuration info? The explorer's script is at
>
> http://classic.sunsolve.sun.com/pub-cgi/show.pl?target=explorer/explorer
>
>> Your opinion will be appreciated. I need to build NFS HA cluster based
>> on multi-tera volumes using SVM so I need to know is there any
>> showstopper.
>>
>>
>>
> Cool.
>
>
> _______________________________________________
> lvm-discuss mailing list
> lvm-discuss at opensolaris dot org
_______________________________________________
lvm-discuss mailing list
lvm-discuss at opensolaris dot org
|
|
|
|
|
Posts:
157
From:
IL
Registered:
6/14/05
|
|
|
|
|
|
Posts:
157
From:
IL
Registered:
6/14/05
|
|
|
|
|
|
Tom Whitten
thomas.whitten@sun.com
|
|
|
|
Re: Re: Re: SVM question: what am I doing wrong?
Posted:
Jan 25, 2006 10:46 AM
in response to: napobo3
|
|
You're URL says "When the MPxIO is enabled, the SVM cannot work with disks
with long device names (64 bytes target name length)." We do, however,
have SVM running with MPXIO and long device names. Can you provide more
details?
tom
Leon Koll writes:
> [i][b]I found the problem!!![/b][/i]
> Looks like a bug in SVM.
> Read here: http://napobo3.blogspot.com/2006/01/i-found-problem.html
> This message posted from opensolaris.org
> _______________________________________________
> lvm-discuss mailing list
> lvm-discuss at opensolaris dot org
_______________________________________________
lvm-discuss mailing list
lvm-discuss at opensolaris dot org
|
|
|
|
|
Leon Koll
napobo3@gmail.com
|
|
|
|
Re: Re: Re: SVM question: what am I doing wrong?
Posted:
Jan 25, 2006 12:38 PM
in response to: Tom Whitten
|
|
Hi Tom,
do you work with device names long like this one:
/dev/dsk/c6t<64bytes>d0s0
What I see that SVM has problems with them.
-- Leon
On 1/25/06, Tom Whitten <thomas dot whitten at sun dot com> wrote:
> You're URL says "When the MPxIO is enabled, the SVM cannot work with disks
> with long device names (64 bytes target name length)." We do, however,
> have SVM running with MPXIO and long device names. Can you provide more
> details?
>
> tom
>
> Leon Koll writes:
> > [i][b]I found the problem!!![/b][/i]
> > Looks like a bug in SVM.
> > Read here: http://napobo3.blogspot.com/2006/01/i-found-problem.html
> > This message posted from opensolaris.org
> > _______________________________________________
> > lvm-discuss mailing list
> > lvm-discuss at opensolaris dot org
>
_______________________________________________
lvm-discuss mailing list
lvm-discuss at opensolaris dot org
|
|
|
|
|
Posts:
480
From:
Registered:
3/9/05
|
|
|
|
Re: Re: Re: SVM question: what am I doing wrong?
Posted:
Jan 25, 2006 12:57 PM
in response to: Leon Koll
|
|
But even this is not completely correct. Since you were able to create
metadevices on that disk when it was not part of the diskset. I am
attaching your email to me regarding this, which I am including below.
To summerize I believe here's how things stand.
1. Without mpxio using the 64 byte vdp one can add disks into the diskset.
2. With mpxio enabled using the 64 byte vdp, adding that disk to the
disk fails.
3. With mpxio enabled, one can create metadevices (stripe in the example
below) on that disk with 64 byte vdp.
4. With 16 byte vdp, everything works fine, i.e. no failures are seen
for the steps described above.
-Sanjay
> Sanjay,
> did I understand you correctly? Do you mean to create a strip on the
> "problematic" disk w/o creating the metaset?
> I did it and it worked:
>
> Single:
> # metainit d10 1 1
> c6t58495620202020204E4558545241202020202020313032353020202020202031d0s0
> d10: Concat/Stripe is setup
>
> Stripe:
> # metainit d20 1 2
> c6t58495620202020204E4558545241202020202020313032353020202020202033d0s0
> c6t58495620202020204E4558545241202020202020313032353020202020202034d0s0
> d20: Concat/Stripe is setup
> # metastat
> d20: Concat/Stripe
> Size: 491520 blocks (240 MB)
> Stripe 0: (interlace: 1024 blocks)
> Device
> Start Block Dbase Reloc
>
> /dev/dsk/c6t58495620202020204E45585452412020202020203130323530202020202020
> 33d0s0 0 No Yes
>
> /dev/dsk/c6t58495620202020204E45585452412020202020203130323530202020202020
> 34d0s0 16384 No Yes
>
> d10: Concat/Stripe
> Size: 33505280 blocks (15 GB)
> Stripe 0:
> Device
> Start Block Dbase Reloc
>
> /dev/dsk/c6t58495620202020204E45585452412020202020203130323530202020202020
> 31d0s0 0 No Yes
>
> Device Relocation Information:
> Device
> Reloc Device ID
> /dev/dsk/c6t58495620202020204E45585452412020202020203130323530202020202020
> 33d0 Yes id1,ssd@TXIV_____NEXTRA______10250______3
> /dev/dsk/c6t58495620202020204E45585452412020202020203130323530202020202020
> 34d0 Yes id1,ssd@TXIV_____NEXTRA______10250______4
> /dev/dsk/c6t58495620202020204E45585452412020202020203130323530202020202020
> 31d0 Yes id1,ssd@TXIV_____NEXTRA______10250______1
>
> Regards,
> -- Leon
>
>
Leon Koll wrote:
> Hi Tom,
> do you work with device names long like this one:
> /dev/dsk/c6t<64bytes>d0s0
> What I see that SVM has problems with them.
>
> -- Leon
>
>
> On 1/25/06, Tom Whitten <thomas dot whitten at sun dot com> wrote:
>
>>You're URL says "When the MPxIO is enabled, the SVM cannot work with disks
>>with long device names (64 bytes target name length)." We do, however,
>>have SVM running with MPXIO and long device names. Can you provide more
>>details?
>>
>>tom
>>
>>Leon Koll writes:
>>
>>>[i][b]I found the problem!!![/b][/i]
>>>Looks like a bug in SVM.
>>>Read here: http://napobo3.blogspot.com/2006/01/i-found-problem.html
>>>This message posted from opensolaris.org
>>>_______________________________________________
>>>lvm-discuss mailing list
>>>lvm-discuss at opensolaris dot org
>>
>>
>>--------------------------------------------------------------------- ---
>>
>>_______________________________________________
>>lvm-discuss mailing list
>>lvm-discuss at opensolaris dot org
_______________________________________________
lvm-discuss mailing list
lvm-discuss at opensolaris dot org
|
|
|
|
|
Leon Koll
napobo3@gmail.com
|
|
|
|
Re: Re: Re: SVM question: what am I doing wrong?
Posted:
Jan 25, 2006 1:38 PM
in response to: nadkarni
|
|
Sanjay,
1. When I wrote ..."SVM has problems with them." (i.e. with long
device names) I meant - it's impossible to make these devices the part
of metaset(s).
In my case (SunCluster+SVM+MPxIO) it means - I cannot work with these
devices at all. In the SunCluster environment the metasets are
mandatory.
2. I agree with your summary except the name "64 byte vdp"
64 bytes it's a target name length - the length of Y in the cXtYdZs0 notation.
To be precise, the evpd page 0x83 INQUIRY page is not 64 bytes long -
in our case it's 40 bytes long:
# /etc/vx/diag.d/vxdmpinq -d -e 1 -p 131 /dev/rdsk/c6t100000062B09FA34d0s2
Inquiry for /dev/rdsk/c6t100000062B09FA34d0s2, evpd 0x1, page code 0x83
/dev/rdsk/c6t100000062B09FA34d0s2: Raw data size 40
Bytes 0 - 9 0x00 0x83 0x00 0x24 0x02 0x01 0x00 0x20 0x58 0x49
Bytes 10 - 19 0x56 0x2d 0x49 0x4e 0x43 0x20 0x58 0x49 0x56 0x31
Bytes 20 - 29 0x30 0x33 0x30 0x20 0x20 0x20 0x20 0x20 0x31 0x30
Bytes 30 - 39 0x31 0x34 0x30 0x20 0x20 0x20 0x20 0x20 0x31 0x33
Regards,
-- Leon
On 1/25/06, Sanjay Nadkarni <Sanjay dot Nadkarni at sun dot com> wrote:
>
> But even this is not completely correct. Since you were able to create
> metadevices on that disk when it was not part of the diskset. I am
> attaching your email to me regarding this, which I am including below.
>
> To summerize I believe here's how things stand.
>
> 1. Without mpxio using the 64 byte vdp one can add disks into the diskset.
>
> 2. With mpxio enabled using the 64 byte vdp, adding that disk to the
> disk fails.
>
> 3. With mpxio enabled, one can create metadevices (stripe in the example
> below) on that disk with 64 byte vdp.
>
> 4. With 16 byte vdp, everything works fine, i.e. no failures are seen
> for the steps described above.
>
> -Sanjay
>
>
> > Sanjay,
> > did I understand you correctly? Do you mean to create a strip on the
> > "problematic" disk w/o creating the metaset?
> > I did it and it worked:
> >
> > Single:
> > # metainit d10 1 1
> > c6t58495620202020204E4558545241202020202020313032353020202020202031d0s0
> > d10: Concat/Stripe is setup
> >
> > Stripe:
> > # metainit d20 1 2
> > c6t58495620202020204E4558545241202020202020313032353020202020202033d0s0
> > c6t58495620202020204E4558545241202020202020313032353020202020202034d0s0
> > d20: Concat/Stripe is setup
> > # metastat
> > d20: Concat/Stripe
> > Size: 491520 blocks (240 MB)
> > Stripe 0: (interlace: 1024 blocks)
> > Device
> > Start Block Dbase Reloc
> >
> > /dev/dsk/c6t58495620202020204E45585452412020202020203130323530202020202020
> > 33d0s0 0 No Yes
> >
> > /dev/dsk/c6t58495620202020204E45585452412020202020203130323530202020202020
> > 34d0s0 16384 No Yes
> >
> > d10: Concat/Stripe
> > Size: 33505280 blocks (15 GB)
> > Stripe 0:
> > Device
> > Start Block Dbase Reloc
> >
> > /dev/dsk/c6t58495620202020204E45585452412020202020203130323530202020202020
> > 31d0s0 0 No Yes
> >
> > Device Relocation Information:
> > Device
> > Reloc Device ID
> > /dev/dsk/c6t58495620202020204E45585452412020202020203130323530202020202020
> > 33d0 Yes id1,ssd@TXIV_____NEXTRA______10250______3
> > /dev/dsk/c6t58495620202020204E45585452412020202020203130323530202020202020
> > 34d0 Yes id1,ssd@TXIV_____NEXTRA______10250______4
> > /dev/dsk/c6t58495620202020204E45585452412020202020203130323530202020202020
> > 31d0 Yes id1,ssd@TXIV_____NEXTRA______10250______1
> >
> > Regards,
> > -- Leon
> >
> >
>
>
> Leon Koll wrote:
> > Hi Tom,
> > do you work with device names long like this one:
> > /dev/dsk/c6t<64bytes>d0s0
> > What I see that SVM has problems with them.
> >
> > -- Leon
> >
> >
> > On 1/25/06, Tom Whitten <thomas dot whitten at sun dot com> wrote:
> >
> >>You're URL says "When the MPxIO is enabled, the SVM cannot work with disks
> >>with long device names (64 bytes target name length)." We do, however,
> >>have SVM running with MPXIO and long device names. Can you provide more
> >>details?
> >>
> >>tom
> >>
> >>Leon Koll writes:
> >>
> >>>[i][b]I found the problem!!![/b][/i]
> >>>Looks like a bug in SVM.
> >>>Read here: http://napobo3.blogspot.com/2006/01/i-found-problem.html
> >>>This message posted from opensolaris.org
> >>>_______________________________________________
> >>>lvm-discuss mailing list
> >>>lvm-discuss at opensolaris dot org
> >>
> >>
> >>----------------------------------------------------------------------- -
> >>
> >>_______________________________________________
> >>lvm-discuss mailing list
> >>lvm-discuss at opensolaris dot org
>
>
_______________________________________________
lvm-discuss mailing list
lvm-discuss at opensolaris dot org
|
|
|
|
|
