We're looking at going the ESX route as well.
The NFS route is thin-provisioned, as you said, which is attractive. Keeping the snapshot capabilities for LUNs by adding space reservations is safer, but costly. Quite frankly, thin provisioning the space res for LUNs is attractive if your organization is pro-active, but that isn't the case where I'm at: too dangerous and asking for danger.
That said, you won't get quite the same latency\throughput with NFS as with iSCSI or especially FCP.
Glenn
-----Original Message----- From: owner-toasters@mathworks.com [mailto:owner-toasters@mathworks.com] On Behalf Of Glenn Dekhayser Sent: Wednesday, July 11, 2007 9:55 AM To: Davies,Matt; Scott Lowe; Forest Leonard; toasters@mathworks.com Subject: RE: List still active?
Advantages IMHO 1) VMDKs on NFS are by default THIN-Provisioned; 2) Throughput is about equal 3) NFS is a heck of a lot easier to administer 4) There's 3x longer history and knowledgebase on NFS network t-shooting.
Dis-Advantages: 1) You can do a QLogic iSCSI card to offload CPU processing of packets; 2) In a big environment you can use iSNS; 3) You can be much more granular with masking/etc (although I can imagine you can be creative with NFS) 4) NFS is stateless (v2), to be run fastest you use UDP; so to make it reliable it gets a little slower (but still not a real biggie)
Anyone else?
Glenn @ VOyant
-----Original Message----- From: owner-toasters@mathworks.com [mailto:owner- toasters@mathworks.com] On Behalf Of Davies,Matt Sent: Wednesday, July 11, 2007 9:35 AM To: Scott Lowe; Forest Leonard; toasters@mathworks.com Subject: RE: List still active?
Scott,
Any chance you could expand on the advantages of NFS over Iscsi ? NFS isn't an area I have any experience of....
Cheers
Matt
-----Original Message----- From: owner-toasters@mathworks.com [mailto:owner- toasters@mathworks.com] On Behalf Of Scott Lowe Sent: 11 July 2007 14:05 To: Forest Leonard; toasters@mathworks.com Subject: Re: List still active?
Forest,
When it comes time to configure VMkernel for VMotion, then I'd definitely recommend keeping it separate from the virtual machine network. As it stands right now, you don't even need a VMkernel NIC configured because it sounds like you are using Qlogic iSCSI HBAs and only have a single ESX Server. Since the Qlogic cards handle the iSCSI traffic and there is no VMotion, there no current need for a VMkernel NIC (unless you want to use NFS from the FAS to provide additional storage for VMs--which, by the way, works pretty well and has some nice advantages over iSCSI, IMHO).
Regards, Scott Lowe ePlus Technology, Inc. slowe@eplus.com
On Jul 11, 2007, at 8:33 AM, Forest Leonard wrote:
Interesting stuff. I only have one ESX server so I haven't gotten to the Vmotion configuration yet.. That should be later this year.
I just configured a RDM to run a Virtual server on. I found an
article
where you want to create the type as NTFS if it is a RDM for a
windows
host. Not sure if I am going to use this going forward. I actually don't know if you can migrate into a RDM. It doesn't look like it.
I am only using 2 NIC's on my Vmware server.. and a QLOGIC card for
my
ISCSI access.. I may need to look at bulking up my NIC configuration. It looks like I may gain some performance if I separate my VMKernal from my Virtual Machine network?
Thanks, Forest
-----Original Message----- From: Davies,Matt [mailto:MDAVIES@generalatlantic.com] Sent: Wednesday, July 11, 2007 8:18 AM To: Forest Leonard; ggwalker@mindspring.com; toasters@mathworks.com Subject: RE: List still active?
Sounds like we are both at the same stage.
We are not using RDM's yet, however when it comes to exchange we
will
have to, or I may just stick with using the Microsoft Iscsi initiatator from within the VM, not exactly supported, but I know of other
people
doing it.
We have migrated 8 machines so far, into just one datastore and have not seen any performance problems at all, although most of the machines have very low IO requirements.
We are using a script to snapshot and then replicate using
snapmirror,
it works very well, however our Virtual Centre server is also a VM which was causing a few problems with the snapshots on the VMware side not being removed, but moving this to a separate datastore seems have cured the problems, even SQL doesn't seem to have a problem.
The script is the one written by Evan Battle, that is in the newest netapp docs on VMware. I did have a few problems with ssh to the filer, but we are now using rsh and it seems to be ok.
I don't know how you have setup your virtual switches on the ESX
side,
but I got some best practice information out of VMware on that subject.
Each of our ESX IBM3550 host servers have 6 Nics, connected as follows.
2 Nics for service console VMKernel for Vmotion, load balance using Virtual port ID
2 Nics for Virtual Machine network, load balance using Virtual port
ID
2 Nics for ISCSI (Service Console and VMKernel), load balance using
IP
Hash
Hope this helps....
-----Original Message----- From: Forest Leonard [mailto:fleonard@rvigroup.com] Sent: 11 July 2007 11:06 To: Davies,Matt; ggwalker@mindspring.com; toasters@mathworks.com Subject: RE: List still active?
Hey Matt... I am actually doing the exact same thing on a FAS 270..
I
have migrated about 7 servers so far.
Are you using RDM's (remote device mappings) for the Virtual Machines? I actually just created 2 200GB LUNs on the netapp to use as
Datastores
and have not had any performance issues.
Just wondering what your experience with RDM's are.. I added one into a virtual machine.. It just lets you map a LUN directly to a Virtual machine.
Has anyone out there used the netapp script to capture a VM
snapshot?
Thanks, Forest
From: owner-toasters@mathworks.com on behalf of Davies,Matt Sent: Wed 7/11/2007 1:00 AM To: ggwalker@mindspring.com; toasters@mathworks.com Subject: RE: List still active?
Still working by the looks of things.
busy in process of migrating all our physical severs to VM, stored on a iscsi lun on a FAS 270.
For those that want to know we are using the software iscsi
initiator
within ESX and have not had any problems so far.
cheers
matt
-----Original Message----- From: owner-toasters@mathworks.com owner-toasters@mathworks.com To: toasters@mathworks.com toasters@mathworks.com Sent: Wed Jul 11 01:03:49 2007 Subject: List still active?
I've noticed that I'm still subscribed, but have received no email since July 4th. Everyone didn't trade their NetApp gear for something
else
while I was out of town did they?? J
This e-mail (including all attachments) is confidential and may be privileged. It is for the exclusive use of the addressee only. If you are not
the
addressee, you are hereby notified that any dissemination of this communication is strictly prohibited. If you have received this communication in error, please erase all copies of the message and its attachments and notify us immediately
at
help@generalatlantic.com mailto:help@generalatlantic.com. Thank
You.
######################################################################
## ############# Note: If you have received this message in error, please notify the
original
sender immediately by telephone (203)975-2100 or by return e-mail,
and
delete the message, along with any attachments from your computer.
If
you have received this message in error, you are hereby notified
that
any dissemination, distribution or copying of this e-mail, and any attachments thereto, is strictly prohibited.This note is also to notify the recipient of this email that it has been scanned for all known viruses and attacking techniques. Thank you. http://www.rvigroup.com
######################################################################
## #############
######################################################################
############### Note: If you have received this message in error, please notify the original sender immediately by telephone (203)975-2100 or by return e-mail, and delete the message, along with any attachments from your computer. If you have received this message in error, you are hereby notified that any dissemination, distribution or copying of this e-mail, and any attachments thereto, is strictly prohibited.This note is also to notify the recipient of this email that it has been scanned for all known viruses and attacking techniques. Thank you. http://www.rvigroup.com
######################################################################
###############
May be I miss something I cannot grasp how NFS and iSCSI (or LUN in general) differ w.r.t. snap reserve. You *must* reserve space for changes, be it on NFS or LUN volume; and amount of reserved space is the same. If you know that you got 20% change, just set fractional reserve to 20. If you do not know amount of change, you must reserve 100% of course for LUN - but then again, you must have 50% free on NFS volume as well.
Where exactly is the difference? Confused thank you.
? ????????? / With best regards / Mit freundlichen Gr??en
--- Andrey Borzenkov Senior system engineer -----Original Message----- From: owner-toasters@mathworks.com [mailto:owner-toasters@mathworks.com] On Behalf Of Glenn Walker Sent: Wednesday, July 11, 2007 6:54 PM To: Glenn Dekhayser; Davies,Matt; Scott Lowe; Forest Leonard; toasters@mathworks.com Subject: RE: List still active?
We're looking at going the ESX route as well.
The NFS route is thin-provisioned, as you said, which is attractive. Keeping the snapshot capabilities for LUNs by adding space reservations is safer, but costly. Quite frankly, thin provisioning the space res for LUNs is attractive if your organization is pro-active, but that isn't the case where I'm at: too dangerous and asking for danger.
That said, you won't get quite the same latency\throughput with NFS as with iSCSI or especially FCP.
Glenn
-----Original Message----- From: owner-toasters@mathworks.com [mailto:owner-toasters@mathworks.com] On Behalf Of Glenn Dekhayser Sent: Wednesday, July 11, 2007 9:55 AM To: Davies,Matt; Scott Lowe; Forest Leonard; toasters@mathworks.com Subject: RE: List still active?
Advantages IMHO 1) VMDKs on NFS are by default THIN-Provisioned; 2) Throughput is about equal 3) NFS is a heck of a lot easier to administer 4) There's 3x longer history and knowledgebase on NFS network t-shooting.
Dis-Advantages: 1) You can do a QLogic iSCSI card to offload CPU processing of packets; 2) In a big environment you can use iSNS; 3) You can be much more granular with masking/etc (although I can imagine you can be creative with NFS) 4) NFS is stateless (v2), to be run fastest you use UDP; so to make it reliable it gets a little slower (but still not a real biggie)
Anyone else?
Glenn @ VOyant
-----Original Message----- From: owner-toasters@mathworks.com [mailto:owner- toasters@mathworks.com] On Behalf Of Davies,Matt Sent: Wednesday, July 11, 2007 9:35 AM To: Scott Lowe; Forest Leonard; toasters@mathworks.com Subject: RE: List still active?
Scott,
Any chance you could expand on the advantages of NFS over Iscsi ? NFS isn't an area I have any experience of....
Cheers
Matt
-----Original Message----- From: owner-toasters@mathworks.com [mailto:owner- toasters@mathworks.com] On Behalf Of Scott Lowe Sent: 11 July 2007 14:05 To: Forest Leonard; toasters@mathworks.com Subject: Re: List still active?
Forest,
When it comes time to configure VMkernel for VMotion, then I'd definitely recommend keeping it separate from the virtual machine network. As it stands right now, you don't even need a VMkernel NIC configured because it sounds like you are using Qlogic iSCSI HBAs and only have a single ESX Server. Since the Qlogic cards handle the iSCSI traffic and there is no VMotion, there no current need for a VMkernel NIC (unless you want to use NFS from the FAS to provide additional storage for VMs--which, by the way, works pretty well and has some nice advantages over iSCSI, IMHO).
Regards, Scott Lowe ePlus Technology, Inc. slowe@eplus.com
On Jul 11, 2007, at 8:33 AM, Forest Leonard wrote:
Interesting stuff. I only have one ESX server so I haven't gotten to the Vmotion configuration yet.. That should be later this year.
I just configured a RDM to run a Virtual server on. I found an
article
where you want to create the type as NTFS if it is a RDM for a
windows
host. Not sure if I am going to use this going forward. I actually don't know if you can migrate into a RDM. It doesn't look like it.
I am only using 2 NIC's on my Vmware server.. and a QLOGIC card for
my
ISCSI access.. I may need to look at bulking up my NIC configuration. It looks like I may gain some performance if I separate my VMKernal from my Virtual Machine network?
Thanks, Forest
-----Original Message----- From: Davies,Matt [mailto:MDAVIES@generalatlantic.com] Sent: Wednesday, July 11, 2007 8:18 AM To: Forest Leonard; ggwalker@mindspring.com; toasters@mathworks.com Subject: RE: List still active?
Sounds like we are both at the same stage.
We are not using RDM's yet, however when it comes to exchange we
will
have to, or I may just stick with using the Microsoft Iscsi initiatator from within the VM, not exactly supported, but I know of other
people
doing it.
We have migrated 8 machines so far, into just one datastore and have not seen any performance problems at all, although most of the machines have very low IO requirements.
We are using a script to snapshot and then replicate using
snapmirror,
it works very well, however our Virtual Centre server is also a VM which was causing a few problems with the snapshots on the VMware side not being removed, but moving this to a separate datastore seems have cured the problems, even SQL doesn't seem to have a problem.
The script is the one written by Evan Battle, that is in the newest netapp docs on VMware. I did have a few problems with ssh to the filer, but we are now using rsh and it seems to be ok.
I don't know how you have setup your virtual switches on the ESX
side,
but I got some best practice information out of VMware on that subject.
Each of our ESX IBM3550 host servers have 6 Nics, connected as follows.
2 Nics for service console VMKernel for Vmotion, load balance using Virtual port ID
2 Nics for Virtual Machine network, load balance using Virtual port
ID
2 Nics for ISCSI (Service Console and VMKernel), load balance using
IP
Hash
Hope this helps....
-----Original Message----- From: Forest Leonard [mailto:fleonard@rvigroup.com] Sent: 11 July 2007 11:06 To: Davies,Matt; ggwalker@mindspring.com; toasters@mathworks.com Subject: RE: List still active?
Hey Matt... I am actually doing the exact same thing on a FAS 270..
I
have migrated about 7 servers so far.
Are you using RDM's (remote device mappings) for the Virtual Machines? I actually just created 2 200GB LUNs on the netapp to use as
Datastores
and have not had any performance issues.
Just wondering what your experience with RDM's are.. I added one into a virtual machine.. It just lets you map a LUN directly to a Virtual machine.
Has anyone out there used the netapp script to capture a VM
snapshot?
Thanks, Forest
From: owner-toasters@mathworks.com on behalf of Davies,Matt Sent: Wed 7/11/2007 1:00 AM To: ggwalker@mindspring.com; toasters@mathworks.com Subject: RE: List still active?
Still working by the looks of things.
busy in process of migrating all our physical severs to VM, stored on a iscsi lun on a FAS 270.
For those that want to know we are using the software iscsi
initiator
within ESX and have not had any problems so far.
cheers
matt
-----Original Message----- From: owner-toasters@mathworks.com owner-toasters@mathworks.com To: toasters@mathworks.com toasters@mathworks.com Sent: Wed Jul 11 01:03:49 2007 Subject: List still active?
I've noticed that I'm still subscribed, but have received no email since July 4th. Everyone didn't trade their NetApp gear for something
else
while I was out of town did they?? J
This e-mail (including all attachments) is confidential and may be privileged. It is for the exclusive use of the addressee only. If you are not
the
addressee, you are hereby notified that any dissemination of this communication is strictly prohibited. If you have received this communication in error, please erase all copies of the message and its attachments and notify us immediately
at
help@generalatlantic.com mailto:help@generalatlantic.com. Thank
You.
######################################################################
## ############# Note: If you have received this message in error, please notify the
original
sender immediately by telephone (203)975-2100 or by return e-mail,
and
delete the message, along with any attachments from your computer.
If
you have received this message in error, you are hereby notified
that
any dissemination, distribution or copying of this e-mail, and any attachments thereto, is strictly prohibited.This note is also to notify the recipient of this email that it has been scanned for all known viruses and attacking techniques. Thank you. http://www.rvigroup.com
######################################################################
## #############
######################################################################
############### Note: If you have received this message in error, please notify the original sender immediately by telephone (203)975-2100 or by return e-mail, and delete the message, along with any attachments from your computer. If you have received this message in error, you are hereby notified that any dissemination, distribution or copying of this e-mail, and any attachments thereto, is strictly prohibited.This note is also to notify the recipient of this email that it has been scanned for all known viruses and attacking techniques. Thank you. http://www.rvigroup.com
######################################################################
###############
The difference is that (when using snapshots) you must reserve 100% of the file. So (as was previously pointed out) for a 50GB LUN, it reserves 50GB + 50GB. As opposed to NFS, where only 50GB would be allocated.
You can certainly set the fractional reserve lower, and this was mentioned earlier. As I pointed out, it can be dangerous if you are not a proactive environment and something slips through the cracks and you fill up your space (EOF = corruption during write). This is no less true with NFS, mind you, but it's a bit of a different mind set I've found.
SO - the upfront reserve for LUNs is higher than NFS ('NFS is thin-provisioned by default' was the statement made), but you can get more 'safety' out of the default implementation.
And we're discussing space reservations, not snap reserve. While similar in concept, these aren't the same thing.
-----Original Message----- From: Borzenkov, Andrey [mailto:Andrey.Borzenkov@fujitsu-siemens.com] Sent: Thursday, July 12, 2007 1:27 AM To: Glenn Walker; Glenn Dekhayser; Davies,Matt; Scott Lowe; Forest Leonard; toasters@mathworks.com Subject: RE: List still active?
May be I miss something I cannot grasp how NFS and iSCSI (or LUN in general) differ w.r.t. snap reserve. You *must* reserve space for changes, be it on NFS or LUN volume; and amount of reserved space is the same. If you know that you got 20% change, just set fractional reserve to 20. If you do not know amount of change, you must reserve 100% of course for LUN - but then again, you must have 50% free on NFS volume as well.
Where exactly is the difference? Confused thank you.
? ????????? / With best regards / Mit freundlichen Gr??en
--- Andrey Borzenkov Senior system engineer -----Original Message----- From: owner-toasters@mathworks.com [mailto:owner-toasters@mathworks.com] On Behalf Of Glenn Walker Sent: Wednesday, July 11, 2007 6:54 PM To: Glenn Dekhayser; Davies,Matt; Scott Lowe; Forest Leonard; toasters@mathworks.com Subject: RE: List still active?
We're looking at going the ESX route as well.
The NFS route is thin-provisioned, as you said, which is attractive. Keeping the snapshot capabilities for LUNs by adding space reservations is safer, but costly. Quite frankly, thin provisioning the space res for LUNs is attractive if your organization is pro-active, but that isn't the case where I'm at: too dangerous and asking for danger.
That said, you won't get quite the same latency\throughput with NFS as with iSCSI or especially FCP.
Glenn
-----Original Message----- From: owner-toasters@mathworks.com [mailto:owner-toasters@mathworks.com] On Behalf Of Glenn Dekhayser Sent: Wednesday, July 11, 2007 9:55 AM To: Davies,Matt; Scott Lowe; Forest Leonard; toasters@mathworks.com Subject: RE: List still active?
Advantages IMHO 1) VMDKs on NFS are by default THIN-Provisioned; 2) Throughput is about equal 3) NFS is a heck of a lot easier to administer 4) There's 3x longer history and knowledgebase on NFS network t-shooting.
Dis-Advantages: 1) You can do a QLogic iSCSI card to offload CPU processing of packets; 2) In a big environment you can use iSNS; 3) You can be much more granular with masking/etc (although I can imagine you can be creative with NFS) 4) NFS is stateless (v2), to be run fastest you use UDP; so to make it reliable it gets a little slower (but still not a real biggie)
Anyone else?
Glenn @ VOyant
-----Original Message----- From: owner-toasters@mathworks.com [mailto:owner- toasters@mathworks.com] On Behalf Of Davies,Matt Sent: Wednesday, July 11, 2007 9:35 AM To: Scott Lowe; Forest Leonard; toasters@mathworks.com Subject: RE: List still active?
Scott,
Any chance you could expand on the advantages of NFS over Iscsi ? NFS isn't an area I have any experience of....
Cheers
Matt
-----Original Message----- From: owner-toasters@mathworks.com [mailto:owner- toasters@mathworks.com] On Behalf Of Scott Lowe Sent: 11 July 2007 14:05 To: Forest Leonard; toasters@mathworks.com Subject: Re: List still active?
Forest,
When it comes time to configure VMkernel for VMotion, then I'd definitely recommend keeping it separate from the virtual machine network. As it stands right now, you don't even need a VMkernel NIC configured because it sounds like you are using Qlogic iSCSI HBAs and only have a single ESX Server. Since the Qlogic cards handle the iSCSI traffic and there is no VMotion, there no current need for a VMkernel NIC (unless you want to use NFS from the FAS to provide additional storage for VMs--which, by the way, works pretty well and has some nice advantages over iSCSI, IMHO).
Regards, Scott Lowe ePlus Technology, Inc. slowe@eplus.com
On Jul 11, 2007, at 8:33 AM, Forest Leonard wrote:
Interesting stuff. I only have one ESX server so I haven't gotten to the Vmotion configuration yet.. That should be later this year.
I just configured a RDM to run a Virtual server on. I found an
article
where you want to create the type as NTFS if it is a RDM for a
windows
host. Not sure if I am going to use this going forward. I actually don't know if you can migrate into a RDM. It doesn't look like it.
I am only using 2 NIC's on my Vmware server.. and a QLOGIC card for
my
ISCSI access.. I may need to look at bulking up my NIC configuration. It looks like I may gain some performance if I separate my VMKernal from my Virtual Machine network?
Thanks, Forest
-----Original Message----- From: Davies,Matt [mailto:MDAVIES@generalatlantic.com] Sent: Wednesday, July 11, 2007 8:18 AM To: Forest Leonard; ggwalker@mindspring.com; toasters@mathworks.com Subject: RE: List still active?
Sounds like we are both at the same stage.
We are not using RDM's yet, however when it comes to exchange we
will
have to, or I may just stick with using the Microsoft Iscsi initiatator from within the VM, not exactly supported, but I know of other
people
doing it.
We have migrated 8 machines so far, into just one datastore and have not seen any performance problems at all, although most of the machines have very low IO requirements.
We are using a script to snapshot and then replicate using
snapmirror,
it works very well, however our Virtual Centre server is also a VM which was causing a few problems with the snapshots on the VMware side not being removed, but moving this to a separate datastore seems have cured the problems, even SQL doesn't seem to have a problem.
The script is the one written by Evan Battle, that is in the newest netapp docs on VMware. I did have a few problems with ssh to the filer, but we are now using rsh and it seems to be ok.
I don't know how you have setup your virtual switches on the ESX
side,
but I got some best practice information out of VMware on that subject.
Each of our ESX IBM3550 host servers have 6 Nics, connected as follows.
2 Nics for service console VMKernel for Vmotion, load balance using Virtual port ID
2 Nics for Virtual Machine network, load balance using Virtual port
ID
2 Nics for ISCSI (Service Console and VMKernel), load balance using
IP
Hash
Hope this helps....
-----Original Message----- From: Forest Leonard [mailto:fleonard@rvigroup.com] Sent: 11 July 2007 11:06 To: Davies,Matt; ggwalker@mindspring.com; toasters@mathworks.com Subject: RE: List still active?
Hey Matt... I am actually doing the exact same thing on a FAS 270..
I
have migrated about 7 servers so far.
Are you using RDM's (remote device mappings) for the Virtual Machines? I actually just created 2 200GB LUNs on the netapp to use as
Datastores
and have not had any performance issues.
Just wondering what your experience with RDM's are.. I added one into a virtual machine.. It just lets you map a LUN directly to a Virtual machine.
Has anyone out there used the netapp script to capture a VM
snapshot?
Thanks, Forest
From: owner-toasters@mathworks.com on behalf of Davies,Matt Sent: Wed 7/11/2007 1:00 AM To: ggwalker@mindspring.com; toasters@mathworks.com Subject: RE: List still active?
Still working by the looks of things.
busy in process of migrating all our physical severs to VM, stored on a iscsi lun on a FAS 270.
For those that want to know we are using the software iscsi
initiator
within ESX and have not had any problems so far.
cheers
matt
-----Original Message----- From: owner-toasters@mathworks.com owner-toasters@mathworks.com To: toasters@mathworks.com toasters@mathworks.com Sent: Wed Jul 11 01:03:49 2007 Subject: List still active?
I've noticed that I'm still subscribed, but have received no email since July 4th. Everyone didn't trade their NetApp gear for something
else
while I was out of town did they?? J
This e-mail (including all attachments) is confidential and may be privileged. It is for the exclusive use of the addressee only. If you are not
the
addressee, you are hereby notified that any dissemination of this communication is strictly prohibited. If you have received this communication in error, please erase all copies of the message and its attachments and notify us immediately
at
help@generalatlantic.com mailto:help@generalatlantic.com. Thank
You.
######################################################################
## ############# Note: If you have received this message in error, please notify the
original
sender immediately by telephone (203)975-2100 or by return e-mail,
and
delete the message, along with any attachments from your computer.
If
you have received this message in error, you are hereby notified
that
any dissemination, distribution or copying of this e-mail, and any attachments thereto, is strictly prohibited.This note is also to notify the recipient of this email that it has been scanned for all known viruses and attacking techniques. Thank you. http://www.rvigroup.com
######################################################################
## #############
######################################################################
############### Note: If you have received this message in error, please notify the original sender immediately by telephone (203)975-2100 or by return e-mail, and delete the message, along with any attachments from your computer. If you have received this message in error, you are hereby notified that any dissemination, distribution or copying of this e-mail, and any attachments thereto, is strictly prohibited.This note is also to notify the recipient of this email that it has been scanned for all known viruses and attacking techniques. Thank you. http://www.rvigroup.com
######################################################################
###############
On 7/12/07, Glenn Walker ggwalker@mindspring.com wrote:
You can certainly set the fractional reserve lower, and this was mentioned earlier. As I pointed out, it can be dangerous if you are not a proactive environment and something slips through the cracks and you fill up your space (EOF = corruption during write). This is no less true with NFS, mind you, but it's a bit of a different mind set I've found.
Are proactive corporate environments not a rarity or am I just jaded.
SJM
That's the problem - there are _some_ but they aren't _here_ where I am :)
-----Original Message----- From: owner-toasters@mathworks.com [mailto:owner-toasters@mathworks.com] On Behalf Of Sheldon Mustard Sent: Thursday, July 12, 2007 2:12 PM To: toasters@mathworks.com Subject: Re: List still active?
On 7/12/07, Glenn Walker ggwalker@mindspring.com wrote:
You can certainly set the fractional reserve lower, and this was mentioned earlier. As I pointed out, it can be dangerous if you are
not
a proactive environment and something slips through the cracks and you fill up your space (EOF = corruption during write). This is no less true with NFS, mind you, but it's a bit of a different mind set I've found.
Are proactive corporate environments not a rarity or am I just jaded.
SJM
I think they exist in fantasyland... :)
-----Original Message----- From: owner-toasters@mathworks.com on behalf of Glenn Walker Sent: Thu 7/12/2007 12:10 PM To: Sheldon Mustard; toasters@mathworks.com Subject: RE: List still active?
That's the problem - there are _some_ but they aren't _here_ where I am :)
-----Original Message----- From: owner-toasters@mathworks.com [mailto:owner-toasters@mathworks.com] On Behalf Of Sheldon Mustard Sent: Thursday, July 12, 2007 2:12 PM To: toasters@mathworks.com Subject: Re: List still active?
On 7/12/07, Glenn Walker ggwalker@mindspring.com wrote:
You can certainly set the fractional reserve lower, and this was mentioned earlier. As I pointed out, it can be dangerous if you are
not
a proactive environment and something slips through the cracks and you fill up your space (EOF = corruption during write). This is no less true with NFS, mind you, but it's a bit of a different mind set I've found.
Are proactive corporate environments not a rarity or am I just jaded.
SJM
I have certainly never worked anywhere where I could count on getting budget approval for storage upgrades any time of year. ("We're about to run out of space on the file server! We need to spend some of our budgeted storage money!" "Sorry, we're in a budget freeze, no capital purchases until next fiscal year...maybe.") And my users always use up all the space they're given rapidly, even when I buy storage based on their 3-year requirements and I double those.
Just one of the reasons I don't allocate space like this. I wish I could though.
To be fair, that's the point of thin provisioning: to save money. But it's that age old battle between time and money... you can spend the time to manage and project and save money, or you can spend money and not have to spend as much time managing and planning.
-----Original Message----- From: owner-toasters@mathworks.com [mailto:owner-toasters@mathworks.com] On Behalf Of Sphar, Mike Sent: Thursday, July 12, 2007 3:46 PM To: toasters@mathworks.com Subject: RE: List still active?
I have certainly never worked anywhere where I could count on getting budget approval for storage upgrades any time of year. ("We're about to run out of space on the file server! We need to spend some of our budgeted storage money!" "Sorry, we're in a budget freeze, no capital purchases until next fiscal year...maybe.") And my users always use up all the space they're given rapidly, even when I buy storage based on their 3-year requirements and I double those.
Just one of the reasons I don't allocate space like this. I wish I could though.
Depends on how you define proactive
a) SysAdmins begging and pleading for more resources before the @#$@#@ hits the fan or b) actually having the resources before the @#$@#$ hits the fan ;-)
Sheldon Mustard wrote:
On 7/12/07, Glenn Walker ggwalker@mindspring.com wrote:
You can certainly set the fractional reserve lower, and this was mentioned earlier. As I pointed out, it can be dangerous if you are not a proactive environment and something slips through the cracks and you fill up your space (EOF = corruption during write). This is no less true with NFS, mind you, but it's a bit of a different mind set I've found.
Are proactive corporate environments not a rarity or am I just jaded.
SJM
May be I miss something I cannot grasp how NFS and iSCSI (or LUN in general) differ w.r.t. snap reserve. You *must* reserve space for changes, be it on NFS or LUN volume; and amount of reserved space is the same. If you know that you got 20% change, just set fractional reserve to 20. If you do not know amount of change, you must reserve 100% of course for LUN - but then again, you must have 50% free on NFS volume as well.
Where exactly is the difference? Confused thank you.
Prior to the recent changes where the active volume can gain space, I saw it as the difference between how NFS clients and LUN clients operated.
When talking to a filesystem (local or NFS), you expect that it might fill. So the NFS clients (and hopefully the application) can handle a temorarily full filesystem.
With a LUN, the situation is different. The OS does not expect that a block write can fail unless there's something wrong with the storage device. So the failure mode from that is probably less defined than a failed NFS write.
-
Borzenkov, Andrey -
Darren Dunham -
Glenn Walker -
Jack Lyons -
Leeds, Daniel -
Sheldon Mustard -
Sphar, Mike