bpierfy
Enthusiast
Enthusiast

AX150i and CX3-20 Questions

Jump to solution

We currently have an AX150i and we need a new SAN for multiple reasons. In looking at the CX3-20, we were told that it is an active/active SAN by Dell and VMware told us it was active/passive. VMware also recommends using MRU.

If the CX3-20 is active passive, why then is "SP Failover" supported, but on the AX150i (which is also active/passive), it is not supported. Maybe I'm not understanding what SP Failover is?

We have 4 nics, and they are currently on 2 vswif's each with two nics teamed. I usually run the SC and the VM networks of vswif0 and the iSCSI stuff off of vswif1.

What am I gaining by going to the CX3-20 with regards to load balancing and failover? In having support for "SP Failover" what do I gain that I didn't have with the AX150i?

0 Kudos
1 Solution

Accepted Solutions
msmenne17
Enthusiast
Enthusiast

bpierfy,

Both the Ax150 and Cx3-20 are Active/Passive arrays.

Only the Symmetix is truly and Active/Active array.

The difference between active/passive and active/active is ownership of the LUNs.

I'm not sure about the AX150 (I think it's the same), but the Cx series arrays have TWO Active/Passive processors. This simply means that a LUN can only be owned by ONE SP at a time. SPA owns LUN0 while SPB owns LUN1. If SPB or the path to SPB dies, then LUN1 will tresspass to SPA and will be owned by SPA now. Just because it is active/passive doesn't mean you're not going to have failover and redundancy.

I had a couple instances with a Cx600 where I lost an SP or a path to an SP. ESX had no trouble at all with the failover. All of my VMs kept running and didn't even register a blip on the radar. The exception to this was the RHEL4 VMs. This wasn't a VMware issue, it was Red Hat issue. Red Hat Entperprise Linux Update 5 supposedly fixes this problem.

Ax/Cx = Active/Passive

Symmetrix = Active/Active

Michael

View solution in original post

0 Kudos
18 Replies
hakimzc
Contributor
Contributor

Are you going to use the iSCSI features of the CX3 or fibre?

If you were to tell VMware that you were going to use iSCSI, they would recommend Active\Passive because that is the only iSCSI configuration they support (and therefore MRU).

Maybe this explains the discrepancy in the information supplied by Dell and VMware?

Respectfully,

bbplc

0 Kudos
bpierfy
Enthusiast
Enthusiast

We are going to use iSCSI.

Interesting, I hadn't though about it that way. Are you telling me that all iSCSI/ESX3 setups are active/passive and MRU?

Message was edited by:

bpierfy (meant active/passive)

0 Kudos
hakimzc
Contributor
Contributor

To be supported, they would have to be.

I don't know if you could get it working in an unsupported fashion in this case.

I was looking at the same SAN, we have a mixture of AX100is and AX150is at the moment. I did a bit of research and decided that for what we wanted, we would be better off with fibre (a bit more expensive I know).

Kind regards,

bbplc

0 Kudos
bpierfy
Enthusiast
Enthusiast

Interestingly, there appears to be some iSCSI SANs that are active/active and support fixed even with software iSCSI initiator. If the CX3-20 is an active/active array (waiting on EMC for answer), why can't ESX use fixed? If its not then that makes sense, see the next question.

I guess my real question is what is "SP failover" exactly? And, the AX150i does not technically do SP failover but the CX3-20 does - what am I gaining from the upgrade?

0 Kudos
hakimzc
Contributor
Contributor

Oh, forgive me I didn't know about those other iSCSI SANs.

By coincidence I am currently involved in a support call with Dell regarding ann AX100i, I'll ask the guy for a definition of SP failover.

Kind regards,

bbplc

0 Kudos
jprior
Enthusiast
Enthusiast

As far as I understand it, SP failover is where the SP with primary ownership of LUNs is unable to communicate with them, and the array fails over to the other SP for pathing.

For example, an ESX host has two FC HBA's, connecting it to both SP A and SP B. LUN0 has SP A as it's primary. Both SP A and SP B have a zone set with SP A, SP B and a FC HBA pathed in the FC switches. Thus when SP A becomes unavailable (h/w fault, IO queue depth exceeded etc.) the array uses SP B to process requests for LUN 0.

The CX3-20 is Active/Active in that they are both up and running - rather than one of them remaining in standby mode), they are not active-active from a VMware perspective - i.e. each controller can present a different set of LUNs to the ESX server but there is only one path to a LUN at any one time. Therefore, to ESX Server they are still active-passive.

I have two CX3-20s (non-ISCSI) each with 3 ESX hosts fiber attached, and different LUNS within my storage group are assigned to different SP's.

0 Kudos
christianZ
Champion
Champion

For clarity,

e.g. EQL works as active/active with "Fixed" path policy (iscsi) although the system has active/standby storage processors.

Active/Active (from vmware site) means, you can access the volumes on that system over both hbas / or storege processors (when you have 2). This way you can balance your load over the 2 hbas (in this case).

More info about ative/passive here:

http://www.vmware.com/community/thread.jspa?threadID=67023&start=15&tstart=0

bpierfy
Enthusiast
Enthusiast

What confuses me about these two arrays (the AX and the CX) is that the AX is supposedly active/passive and doesn't support SP Failover. But we have experienced SP outages (don't ask, it was bad...) and our VMs kept running. I think this is due to the fact that we have 4 paths from each host to each datastore. Granted only one is active at a time, but I think that is always the case.

What about this scenario - Only two paths, but both paths to a single proc (A). Storage proc A fails, and even though we have no path (active or standy) to be, the array still manages to process requests bound for A, through B. Whereas in our case we had paths to each so it didn't matter. If this was the case, we wouldn't need 4 paths - just 2 - one for each NIC of the storage proc in order to provide failover for network problems. Am I on the right track?

If I'm still correct (which would be a miracle at this point), our previous SP failovers would not have been successful if we didn't have a VMKernel and and SC for each storage proc x 2.

One additional piece of information that may or may not help is that each proc on the AX has two NICs and each is place into a different storage VLAN.

0 Kudos
christianZ
Champion
Champion

SP failover supported means by vmware, they are tested it and verified it works under many circumstances.

By AX it can work too - but probably there were problems with it in any situation.

I would personally get the faster one i.e. the CX box - when you get the AX box you won't get more throughput than you have now.

Just my opinion.

0 Kudos
glynnd1
Expert
Expert

For the Clariion range they are all active/passive, the Symmetrix is active/active.

0 Kudos
bpierfy
Enthusiast
Enthusiast

Hopefully this is the last question - I agree that the CX is the way to go (at minimum), but what would we gain by looking into another brand that has something that suits our needs, but is active/active?

As best I can tell from this thread and other research, I'll just gain the ability to do fixed path manual load balancing. I'm also hopeful that someday they can do automatic load balancing, and I'd guess it would require active/active (total speculation on my part).

Thanks (its gonna be hard to award points on this one!)

0 Kudos
Steve_Mc
Hot Shot
Hot Shot

If you go with the CX3-20c (combo) you will get both Fibre and iSCSI ports for connectivity.

A single host can only be connected via 1 or the other.

All CLARiiON (FC, CX and CX3-Series are Active/Passive). This means a LUN is owned by 1 or the other Storage Processor (SP).

Earlier response in this thread described A/A as each SP can actively own and service LUNs. This is the case with CLARiiON, but is not typically called A/A.

However SP and HBA failover are options - based on the failover software and connectivity used.

For AX-Series there is available a lite version of Powerpath, which only gives limited failover and no load balancing. I believe you are speaking of VMware attach in this case thouhg and in that case the failover is native to ESX.

CX3-20 is a pretty sweet choice.

Hope that helps,

Steve

femialpha
Enthusiast
Enthusiast

Powerpath for the AX150 actually does loadbalancing. I have 3 of them! As for other options for a SAN, the only thing you will gain over the clariion is ease of use. Products that implement storage virtualization will be much easier to manage especially if your shop has not SAN experience.

0 Kudos
msmenne17
Enthusiast
Enthusiast

bpierfy,

Both the Ax150 and Cx3-20 are Active/Passive arrays.

Only the Symmetix is truly and Active/Active array.

The difference between active/passive and active/active is ownership of the LUNs.

I'm not sure about the AX150 (I think it's the same), but the Cx series arrays have TWO Active/Passive processors. This simply means that a LUN can only be owned by ONE SP at a time. SPA owns LUN0 while SPB owns LUN1. If SPB or the path to SPB dies, then LUN1 will tresspass to SPA and will be owned by SPA now. Just because it is active/passive doesn't mean you're not going to have failover and redundancy.

I had a couple instances with a Cx600 where I lost an SP or a path to an SP. ESX had no trouble at all with the failover. All of my VMs kept running and didn't even register a blip on the radar. The exception to this was the RHEL4 VMs. This wasn't a VMware issue, it was Red Hat issue. Red Hat Entperprise Linux Update 5 supposedly fixes this problem.

Ax/Cx = Active/Passive

Symmetrix = Active/Active

Michael

0 Kudos
Steve_Mc
Hot Shot
Hot Shot

We were talking about the AX150i (iSCSI verison). I was trying not to be too specific but there have been several Powerpath options over the years.

The first Powerpath verison used on AX100i and AX150i was Powerpath 1.1 iSCSI. This is one example of a Powerpath version with limited capability.

Steve

0 Kudos
Schorschi
Expert
Expert

You can use no pathing software with ESX. PowerPath is not supported.

0 Kudos
Steve_Mc
Hot Shot
Hot Shot

Agreed, I said that 2 threads ago. original quesiton was related to which array to go for I believe. ESX isn't usually the only server attached to the array.

Steve

0 Kudos
opbz
Hot Shot
Hot Shot

Hi,

It actually gets a bit messier.

As of flare 26 the Clariions will be able to work in active/active same as symmetrics. This will not apply to AX150.

Problem is that vmware still will be recommending setting path policy as MRU.

0 Kudos