EVA Upgrade

[Balakrishna Kotian]

Hi, Can some one provide me any documentation on upgrading EVA 3000 to EVA 6000 and currently EVA3000 is in production. Does I need to change only controllers or the complete EVA.
Secondly I would also like to know M5314 drive enclosure has 2 backplane or a single backplane. Would be great if I can get any specification details on M5314.

BR

Bala

[Uwe Zessin]

Haven't seen any documentation - as far as I can tell, an in-place upgrade is not possible, yet.

Apart from the controllers, each M5314 enclosure must be upgraded to M5314A (new I/O modules and EMU). And the new firmware would need to change the internal meta-data.

I strongly recommend that you discuss this with a real expert from HP who assists you in planning the migration. Depending on your current environment, it is very likely that you need to make changes on the server side, too, because the EVA6000 works a bit different than the 3000 and old software might get confused or simply does not recognize the 6000.

[Balakrishna Kotian]

Hi Uwe,

We want to add additional disk enclosure for having more space. One option I see is to upgrade the controller either to EVA5000 or EVA6000. With EVA6000 inplace upgrade is not possible and how about going for EVA5000 in this case is it possible for inplace upgrade. The upgrade should be done without touch the data and just change the controller and add more exclosure.
Second option, can it be possible we replace all 146Gb existing disk to 300 Gb disk without bringing down the EVA.

I would also like to know do u have any knowledge about how many backplane is there in M5314 is it single with both controller attached or dual with individual controllers attached to each backplane.

Thank you for u r earlier reply.

BR
Bala

[Steven Clementi]

Bala:

Last I was told, the only way to Upgrade an EVA from 3/5k to 4/6/8k is to let HP do it AND HP was not performing ANY upgrades until after the initial release of the new EVA's with certain exceptions of course.

I do not thing there is ANY documenttion release publicly yet at all. Internally maybe, but I would have heard about it already from certain sources.

Basically, as Uwe states, the controller, the 5314 EMU's and the metadata on the disks are the essential upgrades.

Asside from this, you will need to upgrade your mutlipath software to Full Featured MPIO (which will work along side Secure Path) (also assuming your runnin Windows, I am not an expert in any other OS so you will need to rely on others for other OS support).

Since you have a 3000, you either have a 2C2D or a 2C3D or a 2C4D. Which is it? The EVA3000 can support up to 4 5314 disk shelves.

Upgrading to a 5000 should be possible.

Upgrading to 300 GB disks should also be possible while the EVA is Online. You would need some extra disk slots to start the process and it would take a while to complete as you will be adding disks... leveling... removing disks... leveling.... adding disks... leveling... removing disks... leveling...and so on and so forth.

As for the backplane, I believe the new 5314R is dual with both controllers being able to access both "channels".

On the 5314, I am not sure.



Steven

[Balakrishna Kotian]

Hi Steven,

Thanks a lot for the information you have provided. We got 2C4D filled with 146Gb disk and now need more space at least 2-3TB so was thinking what are available options. So if we have to increase the space without having a long down time either we can go for changing controller to 5000 or play round adding higher capacity disk but then no free slots. Can un-plug 2 disk as we have level1 set and then add 300Gb hdd and level, add-level, add-level what do u say.

Can I get any detailed info on M5314 enclosure would be great as I wanted to confirm all the 14 disk in an enclosure is connected to both controller. If I loose backplane then my whole enclosure is down or is it partial. I dont think it is called as 5314R but the new exclosures for 4k/6k/8k is 5314A.

Thank you very much for Steven and Uwe for giving me your feedback.

BR
Bala

[Uwe Zessin]

All disk drives in all M5x14 disk drive enclosures are connected via two ports and both controllers had dual-port access to all disks from day 1. (I have seen more than one loop failure, beleive me ;-)

Physically, there is a single printed circuit board (PCB) in the disk drive enclosure, because power and both data lines go through a single 40-pin connector.

The cabeling for the first back-end loop goes from one controller to the first disk drive enclosure, jumps over all disks via the internal backplane, leaves the first disk drive enclosure and goes into the second enclosure and so on. Finally it ends at the second controller.

Cabeling for the second loop starts at the second controller and goes to the first(!) disk drive enclosure - just the other way round. The reason is that without loop switches, both controllers are still able to access the remaining disk drive enclosures should one fail completely.

But a small EVA very likely will not be able to tolerate the total loss of a disk drive enclosure unless you are strictly using VRAID-1 virtual disks.

[Steven Clementi]

Yes... "A".. not "R".

In the case of swapping disks...

If you have NO available space at all, then you can drop your protection level, which is not the greatest thing to do, but it is only temp OR just do one disk at a time since you have protection from disk failure (or simulated failure)

Then remove 2 disks and add the 300's, level...remove 2 more..add 2 more..level.. etc. (as you say).

Once you have enough space, you can remove more disks to replace them as well as reset your protection level before you finish.

How are your disks set up now? 1 Disk Group? 2? more?


Steven

[Balakrishna Kotian]

Hi Uwe, When you say more than one loop failure I cant understand how exactly the failure happens.
Thanks a lot for your information about the PCB in the enclosure. Can I get any documentation on this PCB or enclosure.

Cheers,

Bala

[Balakrishna Kotian]

Hi Steven, If I drop my protection level then if 1-2 disk fails in EVA then my whole EVA is down and I will loose everything. Then I should take the longest method to upgrade all the disk by replacing one by one to be on safer side. Currently we have 1 disk group and this is what HP best practise suggest to have one DG with all disk.

BR
Bala

[Uwe Zessin]

Hello Bala,

by 'more than one loop failure' I did not mean that it happened on one array at the same time. I have seen it on different loops at different times and on different arrays.


It is highly unlikely that HP will hand out any technical information about such details like PCBs.

You can easily see the PCB if you remove a few disk drives from the enclosure ;-)

Just kidding - see the attached picture...

[Steven Clementi]

Bala:

Depends on if you have any available space. If you do not have any available space, you will not be able to swap out a drive since your protection level will then be based on the larger sized disk.

If you have sufficient space to cover the new protection level... Single Protection = 600GB, Double would be 1.2TB, then you can fail a drive and replace it and let it level. Repeat process until you have enough space where you can pull out more than 1 drive, then you can do 2 at a time.

If you have no available space, then you will need to drop your protection level temporarily


Steven

[Balakrishna Kotian]

Hi Uwe / Steven,

Sorry for delay in replying. I got another doubt as of on the Disk enclosure. Like we got 2C4D what happens if the backplane fails in one of the enclosure.

In 2 situtation, one 50% of disk slots are filled and second EVA is full with all disk so single slot free. Is there any possiblity of loosing our data in any of the situation mentioned above.

BR

Bala

[Uwe Zessin]

It's not that easy to generally predict when you get data loss, because the EVA uses a pretty complex data allocation mechanism.

The distribution of the data and the ability to withstand the failure of a complete disk drive enclosure can even change over time if you do a few disk group / ungroup operations.

You really need to check the disk group's "RSS disk state".

[Balakrishna Kotian]

Hi Uwe,

How do I check the RSS disk state and find out will the EVA (loss of data) fail if one of my Enclosure fails.

BR
Bala

[Uwe Zessin · Accepted solution]

It's on the disk group properties page (on the right).

none - means that the system cannot tolerate the failure of a disk drive enclosure, but it does not tell where the limitation is

mirror - means the system can tolerate the failure of a disk drive enclosure as long as you only use VRAID-1 virtual disks

parity - means the system can tolerate the failure of a disk drive enclosure for VRAID-1 or VRAID-5 virtual disks


Note that this is a per disk group characteristic - another group that shares the same disk drive enclosures can have a different state!!

[Balakrishna Kotian]

Hi Uwe,

Thank you very much for all your replies. Do you have any document on this stating when RSS is set to mirror with mix of Vdisk 1 & 5 then single enclosure failure will loose the EVA.
Even though all the slots are filled in the EVA still cannot survive.

Can I change it to parity without disturbing the disk group or we need to redo everything from scratch.

Bala

[Uwe Zessin]

Hello Bala,

I've checked a few documents, but I didn't find a description - it might be hidden in Command View EVA's online help, though.

It does not matter if all slots are filled or not - it really depends on how the data is layed out over all enclosures. Now, you wrote that you have an EVA3000. That model supports a maximum of 4 disk drive enclosures, so you will *never* be able to bring a disk group into 'parity' state, sorry.

For the EVA6000, you would need 8 enclosures, make sure that disk groups are multiple of 8 disks and even then you might get into a situation that 'parity' is not possible.

[Peter Mattei]

Hi Bala

Instead of messing around with upgrading to EVA5k or 6k, why dont you just add another EVA to your environment. This can be done concurrently, will increase the overall performance, does not take much time and you will not put your data on risk!

Cheers
Peter

[Balakrishna Kotian]

Hi Peter,

I had two questions in mind one was upgrade for which I got the answer and second was failure of one enclosure in eva 3000 will bring down my SAN? Because if there is failure and we need to have more then 4 enclosure then we can plan for upgrade of controller or else plan for upgrading disk in the EVA.
As per Mr. Uwe he mentioned in RSS if disk group contains both Vdisk 1 & 5 and set to mirror will cause a failure. If thats the case then my DG is set to mirror so wanted to confirm if its anywhere documented. Once I have that then we can come in to some conclusion which path is required to increase the capacity with complete redundancy.

BR
Bala

[Peter Mattei]

Bala

First you have to understand that the EVA is a midrange array and is designed without SPOF (Single Point of Failure)
This is achieved by having
- 2 Controller
- RAID protection
- Disk enclosures with 2 IO modules and seperate FC loops
- two independend power supplies per enclosure

The loss of an enclosure is not considered a single failure, because it needs more than one component to fail. This could be
- 2 power supplies in an enclosure
- 2 IO modules in an enclosure
- both FC loops in an enclosure

If you want to read more about EVA design considerations like RSS see
http://h20000.www2.hp.com/bc/docs/support/SupportManual/lpg29448/lpg29448.pdf

Cheers
Peter

[Balakrishna Kotian]

Hi Peter,

I very much agree with you choice of option but what I was looking like what happens if the backplance fails in an enclosure which is very rare or multi-disk failure (meltdown) brings down the full enclosure. I also checked the current RSS state is set to mirror and we already got Vdisk1 and Vdisk5 in a single disk group.
As per Mr. Uwe RSS only support Vdisk1 in mirror state and I also used EVA sizer in which when I feed in my existing details and I see RSS dont support. But as soon as I make it to 2C12D the mix Vdisk works.

Any suggestion on this as our EVA3000 is currently in production so that I cannot test failing of one enclosure :) or else I would be the first one to test.

Uwe, I also tried looking to the help menu of EVA still failed to get the piece of information which you have provided. I would appreciate if you help me in getting that part of information in some kind of doc.

I have also attached the screen shot of EVA sizer tool.

BR

Bala

[Balakrishna Kotian]

Uwe,

If I further look more into details of RSS for 29 nos of disk, RSS member will be created as 8+8+8+5 or 8+8+6+7 ? which one is supported or will it work in real scenario.

Tnx

Bala

[Steven Clementi]

Bala:

I believe you will NEVER have an RSS group of 5 disks. RSS strives to be at 8 disks, but depending on the amount of disks you actually have, you can have 6 to 11 disks in 1 RSS.

It is likely that you will have.. 8 + 8 + 7 + 6.

The way it works out is... the RSS starts out at 6 disks and will expand up to 11. Once it hits 12, it splits into 2 groups of 6. When you add more disks, the first group of 6 will increase to 8 disks, then the second group will increase to 8, then if you add more, the last groupd will increase to 11 again and start the process over. (all assuming you adding 1 or 2 disks at a time).

Steven

[Balakrishna Kotian]

Hi Everybody, Thank you for everyones replies and suggestions. I got lots of answers whatever I wanted.
Thank you guys!!

BR
Bala

[Balakrishna Kotian]

Upgrading to EVA 6000 from EVA3000 ;-)