Hi all,
We are considering a FAS3220HA running clustered ONTAP (8.2) with a single DS4246 shelf. However, we were recently informed by our sales people that this is not possible, and that the minimum number of shelves for an HA pair running cDOT was two. Apparently each node needs it's own DS4246 for the cDOT operating system? Is it not possible to split the disks on a single shelf into two aggregates and assign one aggregate to each controller?
Confused, Scott
Scott> We are considering a FAS3220HA running clustered ONTAP (8.2) Scott> with a single DS4246 shelf. However, we were recently informed Scott> by our sales people that this is not possible, and that the Scott> minimum number of shelves for an HA pair running cDOT was Scott> two. Apparently each node needs it's own DS4246 for the cDOT Scott> operating system? Is it not possible to split the disks on a Scott> single shelf into two aggregates and assign one aggregate to Scott> each controller?
Each head needs it's own boot disks, and since it's hard-coded to boot off slots 1 and 2 (or some other pair, I forget) in the shelf, having just one shelf would not allow the other head to boot.
Personally, I'd probably gamble that a single 3220 with a single shelf doesn't really need clustering since you're obviously (to me at least, please correct me if I'm wrong) not looking for capacity or even massive performance, since you have a small number of spindles. But I could be totally off base here.
John
It's all done with software disk ownership, it could boot off of disk 23 or 24 if you wanted to create the aggregates that way. (whatever the slot/bay naming scheme is these days). I know that I could /make it work/ just fine with 2 controllers and one shelf, but it may not be a "supported configuration" from a netapp perspective.
Matt
On Sep 25, 2013, at 1:15 PM, "John Stoffel" john.stoffel@taec.toshiba.com wrote:
Scott> We are considering a FAS3220HA running clustered ONTAP (8.2) Scott> with a single DS4246 shelf. However, we were recently informed Scott> by our sales people that this is not possible, and that the Scott> minimum number of shelves for an HA pair running cDOT was Scott> two. Apparently each node needs it's own DS4246 for the cDOT Scott> operating system? Is it not possible to split the disks on a Scott> single shelf into two aggregates and assign one aggregate to Scott> each controller?
Each head needs it's own boot disks, and since it's hard-coded to boot off slots 1 and 2 (or some other pair, I forget) in the shelf, having just one shelf would not allow the other head to boot.
Personally, I'd probably gamble that a single 3220 with a single shelf doesn't really need clustering since you're obviously (to me at least, please correct me if I'm wrong) not looking for capacity or even massive performance, since you have a small number of spindles. But I could be totally off base here.
John
Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
With so few disks and presumably needing HA, I would almost wonder if you would see better performance creating a 3 disk agg0/vol0 on toaster B side, and adding all the remaining disks to single aggr0 on the A side (housing vol0 and data volumes)
...I know that might seem a little counter intuitive from some perspectives, but there are not many good configuration options, in terms of raid group/aggregate layout , with so few disks.
-----Original Message----- From: toasters-bounces@teaparty.net [mailto:toasters-bounces@teaparty.net] On Behalf Of Matt Hallmark Sent: Wednesday, September 25, 2013 4:21 PM To: John Stoffel Cc: toasters@teaparty.net Subject: Re: What is the minimum number of shelves for a FAS3220 HA pair?
It's all done with software disk ownership, it could boot off of disk 23 or 24 if you wanted to create the aggregates that way. (whatever the slot/bay naming scheme is these days). I know that I could /make it work/ just fine with 2 controllers and one shelf, but it may not be a "supported configuration" from a netapp perspective.
Matt
On Sep 25, 2013, at 1:15 PM, "John Stoffel" john.stoffel@taec.toshiba.com wrote:
Scott> We are considering a FAS3220HA running clustered ONTAP (8.2) Scott> with a single DS4246 shelf. However, we were recently informed Scott> by our sales people that this is not possible, and that the Scott> minimum number of shelves for an HA pair running cDOT was two. Scott> Apparently each node needs it's own DS4246 for the cDOT Scott> operating system? Is it not possible to split the disks on a Scott> single shelf into two aggregates and assign one aggregate to Scott> each controller?
Each head needs it's own boot disks, and since it's hard-coded to boot off slots 1 and 2 (or some other pair, I forget) in the shelf, having just one shelf would not allow the other head to boot.
Personally, I'd probably gamble that a single 3220 with a single shelf doesn't really need clustering since you're obviously (to me at least, please correct me if I'm wrong) not looking for capacity or even massive performance, since you have a small number of spindles. But I could be totally off base here.
John
Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
_______________________________________________ Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
I would simply create a 10-disk aggregate on each head. Carve out an appropriate sized root volume. leave two spares on each head. If you are feeling adventurous, leave one spare on each head.
--tmac
*Tim McCarthy* *Principal Consultant*
Clustered ONTAP Clustered ONTAP NCDA ID: XK7R3GEKC1QQ2LVD RHCE6 110-107-141https://www.redhat.com/wapps/training/certification/verify.html?certNumber=110-107-141&isSearch=False&verify=Verify NCSIE ID: C14QPHE21FR4YWD4 Expires: 08 November 2014 Current until Aug 02, 2016 Expires: 08 November 2014
On Wed, Sep 25, 2013 at 4:28 PM, Jordan Slingerland < Jordan.Slingerland@independenthealth.com> wrote:
With so few disks and presumably needing HA, I would almost wonder if you would see better performance creating a 3 disk agg0/vol0 on toaster B side, and adding all the remaining disks to single aggr0 on the A side (housing vol0 and data volumes)
...I know that might seem a little counter intuitive from some perspectives, but there are not many good configuration options, in terms of raid group/aggregate layout , with so few disks.
-----Original Message----- From: toasters-bounces@teaparty.net [mailto:toasters-bounces@teaparty.net] On Behalf Of Matt Hallmark Sent: Wednesday, September 25, 2013 4:21 PM To: John Stoffel Cc: toasters@teaparty.net Subject: Re: What is the minimum number of shelves for a FAS3220 HA pair?
It's all done with software disk ownership, it could boot off of disk 23 or 24 if you wanted to create the aggregates that way. (whatever the slot/bay naming scheme is these days). I know that I could /make it work/ just fine with 2 controllers and one shelf, but it may not be a "supported configuration" from a netapp perspective.
Matt
On Sep 25, 2013, at 1:15 PM, "John Stoffel" john.stoffel@taec.toshiba.com wrote:
Scott> We are considering a FAS3220HA running clustered ONTAP (8.2) Scott> with a single DS4246 shelf. However, we were recently informed Scott> by our sales people that this is not possible, and that the Scott> minimum number of shelves for an HA pair running cDOT was two. Scott> Apparently each node needs it's own DS4246 for the cDOT Scott> operating system? Is it not possible to split the disks on a Scott> single shelf into two aggregates and assign one aggregate to Scott> each controller?
Each head needs it's own boot disks, and since it's hard-coded to boot off slots 1 and 2 (or some other pair, I forget) in the shelf, having just one shelf would not allow the other head to boot.
Personally, I'd probably gamble that a single 3220 with a single shelf doesn't really need clustering since you're obviously (to me at least, please correct me if I'm wrong) not looking for capacity or even massive performance, since you have a small number of spindles. But I could be totally off base here.
John
Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
Yeah, I guess that would be the obvious fall back. I certainly wasn't suggest anyone follow my advice without thoroughly considering all the pros and cons.
However, with only 8 data disks, I wouldn't expect to see much more than about 1200 w/ops (assuming SATA) sustained, ever.
From: tmac [mailto:tmacmd@gmail.com] Sent: Wednesday, September 25, 2013 4:53 PM To: Jordan Slingerland Cc: Matt Hallmark; John Stoffel; toasters@teaparty.net Subject: Re: What is the minimum number of shelves for a FAS3220 HA pair?
I would simply create a 10-disk aggregate on each head. Carve out an appropriate sized root volume. leave two spares on each head. If you are feeling adventurous, leave one spare on each head.
--tmac
Tim McCarthy Principal Consultant
[http://dl.dropbox.com/u/6874230/na_cert_dma_2c.jpg] [http://dl.dropbox.com/u/6874230/rhce.jpeg] [http://dl.dropbox.com/u/6874230/na_cert_ie-san_2c.jpg]
Clustered ONTAP Clustered ONTAP NCDA ID: XK7R3GEKC1QQ2LVD RHCE6 110-107-141https://www.redhat.com/wapps/training/certification/verify.html?certNumber=110-107-141&isSearch=False&verify=Verify NCSIE ID: C14QPHE21FR4YWD4 Expires: 08 November 2014 Current until Aug 02, 2016 Expires: 08 November 2014
On Wed, Sep 25, 2013 at 4:28 PM, Jordan Slingerland <Jordan.Slingerland@independenthealth.commailto:Jordan.Slingerland@independenthealth.com> wrote: With so few disks and presumably needing HA, I would almost wonder if you would see better performance creating a 3 disk agg0/vol0 on toaster B side, and adding all the remaining disks to single aggr0 on the A side (housing vol0 and data volumes)
...I know that might seem a little counter intuitive from some perspectives, but there are not many good configuration options, in terms of raid group/aggregate layout , with so few disks.
-----Original Message----- From: toasters-bounces@teaparty.netmailto:toasters-bounces@teaparty.net [mailto:toasters-bounces@teaparty.netmailto:toasters-bounces@teaparty.net] On Behalf Of Matt Hallmark Sent: Wednesday, September 25, 2013 4:21 PM To: John Stoffel Cc: toasters@teaparty.netmailto:toasters@teaparty.net Subject: Re: What is the minimum number of shelves for a FAS3220 HA pair?
It's all done with software disk ownership, it could boot off of disk 23 or 24 if you wanted to create the aggregates that way. (whatever the slot/bay naming scheme is these days). I know that I could /make it work/ just fine with 2 controllers and one shelf, but it may not be a "supported configuration" from a netapp perspective.
Matt
On Sep 25, 2013, at 1:15 PM, "John Stoffel" <john.stoffel@taec.toshiba.commailto:john.stoffel@taec.toshiba.com> wrote:
Scott> We are considering a FAS3220HA running clustered ONTAP (8.2) Scott> with a single DS4246 shelf. However, we were recently informed Scott> by our sales people that this is not possible, and that the Scott> minimum number of shelves for an HA pair running cDOT was two. Scott> Apparently each node needs it's own DS4246 for the cDOT Scott> operating system? Is it not possible to split the disks on a Scott> single shelf into two aggregates and assign one aggregate to Scott> each controller?
Each head needs it's own boot disks, and since it's hard-coded to boot off slots 1 and 2 (or some other pair, I forget) in the shelf, having just one shelf would not allow the other head to boot.
Personally, I'd probably gamble that a single 3220 with a single shelf doesn't really need clustering since you're obviously (to me at least, please correct me if I'm wrong) not looking for capacity or even massive performance, since you have a small number of spindles. But I could be totally off base here.
John
Toasters mailing list Toasters@teaparty.netmailto:Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
_______________________________________________ Toasters mailing list Toasters@teaparty.netmailto:Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
_______________________________________________ Toasters mailing list Toasters@teaparty.netmailto:Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
Does not cDOT need separate root aggregate (not volume)?
Отправлено с iPhone
26.09.2013, в 1:04, "tmac" <tmacmd@gmail.commailto:tmacmd@gmail.com> написал(а):
I would simply create a 10-disk aggregate on each head. Carve out an appropriate sized root volume. leave two spares on each head. If you are feeling adventurous, leave one spare on each head.
--tmac
Tim McCarthy Principal Consultant
[http://dl.dropbox.com/u/6874230/na_cert_dma_2c.jpg] [http://dl.dropbox.com/u/6874230/rhce.jpeg] [http://dl.dropbox.com/u/6874230/na_cert_ie-san_2c.jpg]
Clustered ONTAP Clustered ONTAP NCDA ID: XK7R3GEKC1QQ2LVD RHCE6 110-107-141https://www.redhat.com/wapps/training/certification/verify.html?certNumber=110-107-141&isSearch=False&verify=Verify NCSIE ID: C14QPHE21FR4YWD4 Expires: 08 November 2014 Current until Aug 02, 2016 Expires: 08 November 2014
On Wed, Sep 25, 2013 at 4:28 PM, Jordan Slingerland <Jordan.Slingerland@independenthealth.commailto:Jordan.Slingerland@independenthealth.com> wrote: With so few disks and presumably needing HA, I would almost wonder if you would see better performance creating a 3 disk agg0/vol0 on toaster B side, and adding all the remaining disks to single aggr0 on the A side (housing vol0 and data volumes)
...I know that might seem a little counter intuitive from some perspectives, but there are not many good configuration options, in terms of raid group/aggregate layout , with so few disks.
-----Original Message----- From: toasters-bounces@teaparty.netmailto:toasters-bounces@teaparty.net [mailto:toasters-bounces@teaparty.netmailto:toasters-bounces@teaparty.net] On Behalf Of Matt Hallmark Sent: Wednesday, September 25, 2013 4:21 PM To: John Stoffel Cc: toasters@teaparty.netmailto:toasters@teaparty.net Subject: Re: What is the minimum number of shelves for a FAS3220 HA pair?
It's all done with software disk ownership, it could boot off of disk 23 or 24 if you wanted to create the aggregates that way. (whatever the slot/bay naming scheme is these days). I know that I could /make it work/ just fine with 2 controllers and one shelf, but it may not be a "supported configuration" from a netapp perspective.
Matt
On Sep 25, 2013, at 1:15 PM, "John Stoffel" <john.stoffel@taec.toshiba.commailto:john.stoffel@taec.toshiba.com> wrote:
Scott> We are considering a FAS3220HA running clustered ONTAP (8.2) Scott> with a single DS4246 shelf. However, we were recently informed Scott> by our sales people that this is not possible, and that the Scott> minimum number of shelves for an HA pair running cDOT was two. Scott> Apparently each node needs it's own DS4246 for the cDOT Scott> operating system? Is it not possible to split the disks on a Scott> single shelf into two aggregates and assign one aggregate to Scott> each controller?
Each head needs it's own boot disks, and since it's hard-coded to boot off slots 1 and 2 (or some other pair, I forget) in the shelf, having just one shelf would not allow the other head to boot.
Personally, I'd probably gamble that a single 3220 with a single shelf doesn't really need clustering since you're obviously (to me at least, please correct me if I'm wrong) not looking for capacity or even massive performance, since you have a small number of spindles. But I could be totally off base here.
John
Toasters mailing list Toasters@teaparty.netmailto:Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
_______________________________________________ Toasters mailing list Toasters@teaparty.netmailto:Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
_______________________________________________ Toasters mailing list Toasters@teaparty.netmailto:Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
_______________________________________________ Toasters mailing list Toasters@teaparty.netmailto:Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
Correct. Each node gets its own root aggr (3 disk min). These are not usable for data unlike 7mode. Then once booted, you create an aggregate for data, and then vservers to use it.
Each Node: 2 disks - spare 3 disks - root aggr 7 disks - data aggregate, vservers can share. You lose 2 parity here so 5 drives worth of data per node.
The $/TB should be high enough to laugh the deal off the table, which is why the 2 shelf scenario comes out I suspect. With 48 drives to play with (assuming even split) you get 17 data 2 parity usable by the vserver(s) on each node and the overhead for the cluster mode is a much smaller percentage of the whole.
Note: If you do a single node CDoT (supported in 8.2) then you'd have a supported 1 shelf config. But at the cost of HA.
Colin Bieberstein
- Sent from my iPad
On Sep 26, 2013, at 12:10 AM, "Borzenkov, Andrey" andrey.borzenkov@ts.fujitsu.com wrote:
Does not cDOT need separate root aggregate (not volume)?
Отправлено с iPhone
26.09.2013, в 1:04, "tmac" tmacmd@gmail.com написал(а):
I would simply create a 10-disk aggregate on each head. Carve out an appropriate sized root volume. leave two spares on each head. If you are feeling adventurous, leave one spare on each head.
--tmac
Tim McCarthy Principal Consultant
Clustered ONTAP Clustered ONTAP
NCDA ID: XK7R3GEKC1QQ2LVD RHCE6 110-107-141 NCSIE ID: C14QPHE21FR4YWD4 Expires: 08 November 2014 Current until Aug 02, 2016 Expires: 08 November 2014
On Wed, Sep 25, 2013 at 4:28 PM, Jordan Slingerland Jordan.Slingerland@independenthealth.com wrote: With so few disks and presumably needing HA, I would almost wonder if you would see better performance creating a 3 disk agg0/vol0 on toaster B side, and adding all the remaining disks to single aggr0 on the A side (housing vol0 and data volumes)
...I know that might seem a little counter intuitive from some perspectives, but there are not many good configuration options, in terms of raid group/aggregate layout , with so few disks.
-----Original Message----- From: toasters-bounces@teaparty.net [mailto:toasters-bounces@teaparty.net] On Behalf Of Matt Hallmark Sent: Wednesday, September 25, 2013 4:21 PM To: John Stoffel Cc: toasters@teaparty.net Subject: Re: What is the minimum number of shelves for a FAS3220 HA pair?
It's all done with software disk ownership, it could boot off of disk 23 or 24 if you wanted to create the aggregates that way. (whatever the slot/bay naming scheme is these days). I know that I could /make it work/ just fine with 2 controllers and one shelf, but it may not be a "supported configuration" from a netapp perspective.
Matt
On Sep 25, 2013, at 1:15 PM, "John Stoffel" john.stoffel@taec.toshiba.com wrote:
Scott> We are considering a FAS3220HA running clustered ONTAP (8.2) Scott> with a single DS4246 shelf. However, we were recently informed Scott> by our sales people that this is not possible, and that the Scott> minimum number of shelves for an HA pair running cDOT was two. Scott> Apparently each node needs it's own DS4246 for the cDOT Scott> operating system? Is it not possible to split the disks on a Scott> single shelf into two aggregates and assign one aggregate to Scott> each controller?
Each head needs it's own boot disks, and since it's hard-coded to boot off slots 1 and 2 (or some other pair, I forget) in the shelf, having just one shelf would not allow the other head to boot.
Personally, I'd probably gamble that a single 3220 with a single shelf doesn't really need clustering since you're obviously (to me at least, please correct me if I'm wrong) not looking for capacity or even massive performance, since you have a small number of spindles. But I could be totally off base here.
John
Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
I did gloss over that...Sorry about that...I am so used to questions regarding 7-mode...
Yep. That will make the disk utilization stink pretty badly. hard sell for c-mode using a single shelf. It can be done. It will not be very performant....
--tmac
*Tim McCarthy* *Principal Consultant*
Clustered ONTAP Clustered ONTAP NCDA ID: XK7R3GEKC1QQ2LVD RHCE6 110-107-141https://www.redhat.com/wapps/training/certification/verify.html?certNumber=110-107-141&isSearch=False&verify=Verify NCSIE ID: C14QPHE21FR4YWD4 Expires: 08 November 2014 Current until Aug 02, 2016 Expires: 08 November 2014
On Thu, Sep 26, 2013 at 2:10 AM, Borzenkov, Andrey < andrey.borzenkov@ts.fujitsu.com> wrote:
Does not cDOT need separate root aggregate (not volume)?
Отправлено с iPhone
26.09.2013, в 1:04, "tmac" tmacmd@gmail.com написал(а):
I would simply create a 10-disk aggregate on each head. Carve out an appropriate sized root volume. leave two spares on each head. If you are feeling adventurous, leave one spare on each head.
--tmac
*Tim McCarthy* *Principal Consultant*
Clustered ONTAP Clustered ONTAP
NCDA ID: XK7R3GEKC1QQ2LVD RHCE6 110-107-141https://www.redhat.com/wapps/training/certification/verify.html?certNumber=110-107-141&isSearch=False&verify=Verify NCSIE ID: C14QPHE21FR4YWD4 Expires: 08 November 2014 Current until Aug 02, 2016 Expires: 08 November 2014
On Wed, Sep 25, 2013 at 4:28 PM, Jordan Slingerland < Jordan.Slingerland@independenthealth.com> wrote:
With so few disks and presumably needing HA, I would almost wonder if you would see better performance creating a 3 disk agg0/vol0 on toaster B side, and adding all the remaining disks to single aggr0 on the A side (housing vol0 and data volumes)
...I know that might seem a little counter intuitive from some perspectives, but there are not many good configuration options, in terms of raid group/aggregate layout , with so few disks.
-----Original Message----- From: toasters-bounces@teaparty.net [mailto:toasters-bounces@teaparty.net] On Behalf Of Matt Hallmark Sent: Wednesday, September 25, 2013 4:21 PM To: John Stoffel Cc: toasters@teaparty.net Subject: Re: What is the minimum number of shelves for a FAS3220 HA pair?
It's all done with software disk ownership, it could boot off of disk 23 or 24 if you wanted to create the aggregates that way. (whatever the slot/bay naming scheme is these days). I know that I could /make it work/ just fine with 2 controllers and one shelf, but it may not be a "supported configuration" from a netapp perspective.
Matt
On Sep 25, 2013, at 1:15 PM, "John Stoffel" < john.stoffel@taec.toshiba.com> wrote:
Scott> We are considering a FAS3220HA running clustered ONTAP (8.2) Scott> with a single DS4246 shelf. However, we were recently informed Scott> by our sales people that this is not possible, and that the Scott> minimum number of shelves for an HA pair running cDOT was two. Scott> Apparently each node needs it's own DS4246 for the cDOT Scott> operating system? Is it not possible to split the disks on a Scott> single shelf into two aggregates and assign one aggregate to Scott> each controller?
Each head needs it's own boot disks, and since it's hard-coded to boot off slots 1 and 2 (or some other pair, I forget) in the shelf, having just one shelf would not allow the other head to boot.
Personally, I'd probably gamble that a single 3220 with a single shelf doesn't really need clustering since you're obviously (to me at least, please correct me if I'm wrong) not looking for capacity or even massive performance, since you have a small number of spindles. But I could be totally off base here.
John
Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
I appreciate all the useful information and advice. I have been browsing the cDOT 8.2 Admin Guide this morning (fascinating stuff BTW) and it seems that by default new cDOT systems ship with a three disk root aggr containing the root vol. Don't mess with it. That's the way it is.
The node's root aggregate contains the node's root volume. Starting with Data ONTAP 8.1, new systems are shipped with the root volume in a dedicated, 64-bit root aggregate that contains three disks. By default, a node is set up to use a hard disk drive (HDD) aggregate for the root aggregate. When no HDDs are available, the node is set up to use a solid-state drive (SSD) aggregate for the root aggregate.
The root aggregate must be dedicated to the root volume only. You must not include or create data volumes in the root aggregate.
I was hoping to implement cDOT now with the purchase of a new system to facilitate easier transitions to new hardware in the future, but with a limited budget and the $$ increment of adding a second shelf to a FAS3220 I may need to reconsider either 7-mode or a FAS2440-4.
IOPS are not our major concern. Our highest priority is capacity and HA. Our current FAS3140 has been great on both of those counts.
Cheers, Scott
On Sep 26, 2013, at 2:48 AM, tmac wrote:
I did gloss over that...Sorry about that...I am so used to questions regarding 7-mode...
Yep. That will make the disk utilization stink pretty badly. hard sell for c-mode using a single shelf. It can be done. It will not be very performant....
--tmac
Tim McCarthy Principal Consultant
Clustered ONTAP Clustered ONTAP
NCDA ID: XK7R3GEKC1QQ2LVD RHCE6 110-107-141 NCSIE ID: C14QPHE21FR4YWD4 Expires: 08 November 2014 Current until Aug 02, 2016 Expires: 08 November 2014
On Thu, Sep 26, 2013 at 2:10 AM, Borzenkov, Andrey andrey.borzenkov@ts.fujitsu.com wrote: Does not cDOT need separate root aggregate (not volume)?
Отправлено с iPhone
26.09.2013, в 1:04, "tmac" tmacmd@gmail.com написал(а):
I would simply create a 10-disk aggregate on each head. Carve out an appropriate sized root volume. leave two spares on each head. If you are feeling adventurous, leave one spare on each head.
--tmac
Tim McCarthy Principal Consultant
Clustered ONTAP Clustered ONTAP
NCDA ID: XK7R3GEKC1QQ2LVD RHCE6 110-107-141 NCSIE ID: C14QPHE21FR4YWD4 Expires: 08 November 2014 Current until Aug 02, 2016 Expires: 08 November 2014
On Wed, Sep 25, 2013 at 4:28 PM, Jordan Slingerland Jordan.Slingerland@independenthealth.com wrote: With so few disks and presumably needing HA, I would almost wonder if you would see better performance creating a 3 disk agg0/vol0 on toaster B side, and adding all the remaining disks to single aggr0 on the A side (housing vol0 and data volumes)
...I know that might seem a little counter intuitive from some perspectives, but there are not many good configuration options, in terms of raid group/aggregate layout , with so few disks.
-----Original Message----- From: toasters-bounces@teaparty.net [mailto:toasters-bounces@teaparty.net] On Behalf Of Matt Hallmark Sent: Wednesday, September 25, 2013 4:21 PM To: John Stoffel Cc: toasters@teaparty.net Subject: Re: What is the minimum number of shelves for a FAS3220 HA pair?
It's all done with software disk ownership, it could boot off of disk 23 or 24 if you wanted to create the aggregates that way. (whatever the slot/bay naming scheme is these days). I know that I could /make it work/ just fine with 2 controllers and one shelf, but it may not be a "supported configuration" from a netapp perspective.
Matt
On Sep 25, 2013, at 1:15 PM, "John Stoffel" john.stoffel@taec.toshiba.com wrote:
Scott> We are considering a FAS3220HA running clustered ONTAP (8.2) Scott> with a single DS4246 shelf. However, we were recently informed Scott> by our sales people that this is not possible, and that the Scott> minimum number of shelves for an HA pair running cDOT was two. Scott> Apparently each node needs it's own DS4246 for the cDOT Scott> operating system? Is it not possible to split the disks on a Scott> single shelf into two aggregates and assign one aggregate to Scott> each controller?
Each head needs it's own boot disks, and since it's hard-coded to boot off slots 1 and 2 (or some other pair, I forget) in the shelf, having just one shelf would not allow the other head to boot.
Personally, I'd probably gamble that a single 3220 with a single shelf doesn't really need clustering since you're obviously (to me at least, please correct me if I'm wrong) not looking for capacity or even massive performance, since you have a small number of spindles. But I could be totally off base here.
John
Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
If you do not need capacity or performance, why not smaller FAS2xxx
A 10 data disks per aggregate is going to be sloooooooow.
-----Original Message----- From: toasters-bounces@teaparty.net [mailto:toasters-bounces@teaparty.net] On Behalf Of John Stoffel Sent: Wednesday, September 25, 2013 4:16 PM To: Scott Classen Cc: toasters@teaparty.net Subject: Re: What is the minimum number of shelves for a FAS3220 HA pair?
Scott> We are considering a FAS3220HA running clustered ONTAP (8.2) with Scott> a single DS4246 shelf. However, we were recently informed by our Scott> sales people that this is not possible, and that the minimum Scott> number of shelves for an HA pair running cDOT was two. Apparently Scott> each node needs it's own DS4246 for the cDOT operating system? Is Scott> it not possible to split the disks on a single shelf into two Scott> aggregates and assign one aggregate to each controller?
Each head needs it's own boot disks, and since it's hard-coded to boot off slots 1 and 2 (or some other pair, I forget) in the shelf, having just one shelf would not allow the other head to boot.
Personally, I'd probably gamble that a single 3220 with a single shelf doesn't really need clustering since you're obviously (to me at least, please correct me if I'm wrong) not looking for capacity or even massive performance, since you have a small number of spindles. But I could be totally off base here.
John
_______________________________________________ Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
I can tell you for sure, it works. I am fairly certain it is supported. Not sure why it would not be. ALL filers today use Software Ownership of disks. All platforms.
If you boot up one unit into maintenance mode: disown all disks. Halt Boot the other node into maintenance mode: disown all disks.
Now, you can use some command to: disk assign xxx 12 where xxx is some combination of disk size, disk type, etc...defining which disks. The 12 is the number to assign. It will pick 12. Then you can do the same (or disk assign all, since 12 would be left)
Then you can configure and go.
I do this now. You must/should turn off the disk auto-assign function.
For the record, the slot 1/2 thing was during the hardware ownership and it read a special section of the disk where the OS was stored.
Todays software ownership systems boots off an internal flash drive only using the disks for storing configuration.
I have a 4-node Clustered Ontap 8.2 system built this way. I started from scratch, zeroed all disks, Attached one shelf to a pair and then assigned disks. I grew them until all the disks were spilt evenly.
Since the disk-auto-assign is disabled and since no head owns all the disks in a shelf, I must assign the disk to head when I replace a failed disk. Not a big deal.
--tmac
*Tim McCarthy* *Principal Consultant*
Clustered ONTAP Clustered ONTAP NCDA ID: XK7R3GEKC1QQ2LVD RHCE6 110-107-141https://www.redhat.com/wapps/training/certification/verify.html?certNumber=110-107-141&isSearch=False&verify=Verify NCSIE ID: C14QPHE21FR4YWD4 Expires: 08 November 2014 Current until Aug 02, 2016 Expires: 08 November 2014
On Wed, Sep 25, 2013 at 4:24 PM, Jordan Slingerland < Jordan.Slingerland@independenthealth.com> wrote:
If you do not need capacity or performance, why not smaller FAS2xxx
A 10 data disks per aggregate is going to be sloooooooow.
-----Original Message----- From: toasters-bounces@teaparty.net [mailto:toasters-bounces@teaparty.net] On Behalf Of John Stoffel Sent: Wednesday, September 25, 2013 4:16 PM To: Scott Classen Cc: toasters@teaparty.net Subject: Re: What is the minimum number of shelves for a FAS3220 HA pair?
Scott> We are considering a FAS3220HA running clustered ONTAP (8.2) with Scott> a single DS4246 shelf. However, we were recently informed by our Scott> sales people that this is not possible, and that the minimum Scott> number of shelves for an HA pair running cDOT was two. Apparently Scott> each node needs it's own DS4246 for the cDOT operating system? Is Scott> it not possible to split the disks on a single shelf into two Scott> aggregates and assign one aggregate to each controller?
Each head needs it's own boot disks, and since it's hard-coded to boot off slots 1 and 2 (or some other pair, I forget) in the shelf, having just one shelf would not allow the other head to boot.
Personally, I'd probably gamble that a single 3220 with a single shelf doesn't really need clustering since you're obviously (to me at least, please correct me if I'm wrong) not looking for capacity or even massive performance, since you have a small number of spindles. But I could be totally off base here.
John
Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters