HPE EVA Storage

EVA 5000 and Current SAN Arrays

 
SOLVED
Go to solution
jhetrick
Advisor

EVA 5000 and Current SAN Arrays

Hi I currently have and EVA 5000 and was wondering if anyone has information on the newer Storage Arrays. My main concern with purchasing another HP is the leveling process takes forever if there is a disk failure.

What I would like to know is on the newer arrays are the Disks all within one huge pool and you carve out your VDisks from that or are they actually separate so that if you have a disk failure and are in the leveling process and you have a second disk failure do you lose the vDisk or the entire array?

Any feedback on this or recommendations would be appreciated!
12 REPLIES 12
Matt Weil
Advisor

Re: EVA 5000 and Current SAN Arrays

leveling and reconstruction of a failed disk are two different things.

leveling your EVA will constantly do this. this is low priority and normal the more data has changed, the bigger the spindle the longer this will take. This is also true with a disk reconstruction.

EVAs do have a RSS (basically RAID 5) group weakness.

if you lose two spindles in a single EVA RSS group you may lose the entire disk group. But losing two disks at a time may or may not mean you lose the disk group. ask your support person if he/she can show these groups.

you can limit your failure domain by making your disk groups smaller. But suffer the raw verses usable overhead of doing this also the benefit of lots of spindles speed.

By newer storage do you mean none HP "virtual"? The same failure problem is there but you failure domain is much smaller. Speed benefit is questionable on the EVA over a.. say 7+1 with a nice LVM.

The EVA's are statistically 5 X 9 boxes but are by no means built this way.

hope this helps

Matt
jhetrick
Advisor

Re: EVA 5000 and Current SAN Arrays

Hi Matt,

When I experience a disk failure my entire EVA goes into a leveling mode that takes 48+ hours...the leveling that is occuring during the disk failure does seem to be excessive. Even if this is a background process...the potential to lose the entire disk group if another disk fails to me is crazy :-) What I have is a very large RAID 5 storage array and that is something I am trying to avoid...that is why I asked if the newer arrays have resolved this issue or do I need to create more disk groups so I have multiple RAID 5's.

I am willing to lose the disk space and spindle speed to a point in order to have a little more reliability than a huge RAID 5 storage array. Unfortunatly for me the person prior to myself set this system up and created 2 disk groups (one for the 72 GB and one for the 146 GB FC Drives) I am now responsible for it.

Thank you for your feedback!!
Uwe Zessin
Honored Contributor
Solution

Re: EVA 5000 and Current SAN Arrays

> you can limit your failure domain by making your disk groups smaller.

No, I don't think so, because the 'preferred' RSS size is 8 disk drives anyway. (I wrote 'preferred' because I've seen many EVAs with RSS sizes of 6,8 and 10 within one disk group at the same time).


jhetrick,
the latest EVA4400/6400/8400 offers VRAID-6.
.
Víctor Cespón
Honored Contributor

Re: EVA 5000 and Current SAN Arrays

I must add that the fear of data loss due to a double failure in a RSS is very unjustified. After handling 3000+ calls of customers with EVAs I do not recall a single case of data loss due to that.

There has been cases of data unavailable for some time due to multiple problems on loops, or controllers, but after replacing hardware, the data is still there.

In any case, as Uwe says, the latest generation of EVAs lets you create LUNs in VRAID 6.
jhetrick
Advisor

Re: EVA 5000 and Current SAN Arrays

Awesome for the RAID 6 thank you so much that does help with the comfort level. I appreciate your time!!
Matt Weil
Advisor

Re: EVA 5000 and Current SAN Arrays

splits and merges of the RSS groups happens automagicly if a disk fails or disks are added to a disk group. Yes ideally the disks and RSS should be added in numbers of 8 disks.

What I meant by failure domain is if you have 50 disks in a group and two disks fail you have lost 50 disks minus the overhead worth of data. It does not matter how the RSS groups are dived. you still lost the data. so if smaller disk groups are used the less amount of data lost can occur.

raid 6 yes more protection can loose 3 disks in a group before data lose but you also take a write penalty due to the extra IOs needed for the double parity.
Uwe Zessin
Honored Contributor

Re: EVA 5000 and Current SAN Arrays

Sounds like you are suggesting to revert back to a 'traditional' RAID-1 array, because you will lose the last amount of data when two mirror disks fail ;-)
.
Matt Weil
Advisor

Re: EVA 5000 and Current SAN Arrays

just giving feedback.

the chances of a double disk failure in a single rss group is small.

but it can happen.

as the spindle sizes get bigger the rebuild time increases.

there are also other failures other than normal disk failures such as shelf, loop, firmware bug or a disk on the loop causing chatter that can cause a double failure.

if something like this happens the controllers will protect the data and kick the disk group offline.

again feedback hope this helps
Uwe Zessin
Honored Contributor

Re: EVA 5000 and Current SAN Arrays

That's fine with me, but do you _really_ set up/suggest to run EVAs with 8-member diskgroups? And what about the 'so-called' "disk protection level"? I am sure you are aware that the space reservation for disk rebuilds is per disk group - or do you suggest to keep some ungroup disks and manually add them after a failure?
.