Equallogic SAN design? Opinions?

[Lanz]

Just curious how other people have setup their Equallogic SANS, I want to hear about various different designs, as everyone seems to have their own ideas about storage.

Here's what we currently have, I've kinda inherited it, so didnt set it up myself.
We have 5 PS6110X and XS members in a group for VMware, inside that group we have a single storage pool with our volumes. They aren't replicated anywhere else and there's no snapshots setup. I'm a little concerned that if we lose one member, then most (if not all volumes) will fail. I've been told by Dell that this is quite normal and the chances of a whole member failing would be extreamly low. We are using Veeam for backups, but it would still take a long time to restore back into a fresh volume if anything happened to the live ones.

How do you have your equallogic's setup, and does anyone have any tips or gotcha's in regards to these SANs?]

cheers

 

[Skidilliplop]

The equalogics can replicate data, but it's not great. They replicate snapshots and on schedule, it's not synchronous and the replica volumes are offline and require manual fail over.

In short, if you want Highly available business continuity (in a disaster scenario) the EQL PS series aren't really the tool for the job.

However I have several and they're not replicated. We've never had a drive failure in going on 3 years let alone a complete member failure. We just accept that if we're THAT unlucky to lose a whole member, we will have to restore it from backup.
You can fudge a bit of resiliency by using storage pools. This would allow you to prevent a logical volume 'roaming' beyond a defined set of members. E.G storage pool 1 contains members 1 & 2, storage pool 2 contains 3 & 4, if 4 fails then volumes in pool 1 are unaffected. But that is a fudge.

You could use 3rd party software to replicate volumes to a DR setup... but when I looked at this the cost of the DR setup didnt' justify it. It was more cost effective to take the downtime hit and restore from backups.

But Dell are right, for the Member to fail in such a way it's un-recoverable is rare. Even if both controllers failed, provided the config is backed up they're FRUs and can just be swapped, config uploaded and back in business. The only scenario that would result in permanent data loss is simultanious multiple drive failure. To get that you'd be REALLY unlucky.

 

[Lanz]

Thanks, thats really useful. Im going to propose a re-design where we only have 2 members in each pool, and we shift the volumes acroess the pools.

I have a couple of these units unused in test lab currently, and looking at them, I think I can change pools, move members and volumes without any downtime. Is that correct?

 

[Skidilliplop]

I would think so, provided there is sufficient storage left in the pool you're removing them from to accommodate all volumes bound to that pool.

This isn't something I've tried doing on mass. If in doubt, talk to their support guys. They're usually pretty good with this sort of stuff.

 

[Lanz]

Any experience with Veeam on Equallogic? (specifically SAN based backup's) ?

 

[Skidilliplop]

Any experience with Veeam on Equallogic? (specifically SAN based backup's) ?

I don't use Veeam in production but I've used trials of it and we use a VMWare aware backup solution for VMs on our SANs which uses similar methods to create backups.
Performance is decent enough. The speed obviously depends on the workload and layout of the storage estate but we've had no problems running backups of Virtual machines off of our EQLs.
Generally speaking the SAN doesn't affect VM backup much as the software is pretty oblivious to it. All it sees are VMDKs and Datastores, and those are all it's really interested in. Veeam couldn't give a hoot what the underlying storage system is.
The only time you might need to worry about SANs and backup software is if you have installed plugins on the hypervisors to do with snapshots, as often a snapshot of the VM is the starting block of the backup process. We don't use that, we don't have any real need to.

 

[Lanz]

Just that I've read, for Veeam to do SAN based backups, the veeam server needs `read only` access to the EQL volumes that have the VMFS on - But in the EQL grpadmin, you cant set access rights on an `per initiator` level, only at a volume level. I'm a little worried that the physical Veeam server has the potential to trash the vmfs disks. I'm probably being too risk adverse, but it would be good to hear of other peoples experiences.

ta

 

[ecksmen]

The equalogics can replicate data, but it's not great. They replicate snapshots and on schedule, it's not synchronous and the replica volumes are offline and require manual fail over.

In short, if you want Highly available business continuity (in a disaster scenario) the EQL PS series aren't really the tool for the job.

However I have several and they're not replicated. We've never had a drive failure in going on 3 years let alone a complete member failure. We just accept that if we're THAT unlucky to lose a whole member, we will have to restore it from backup.
You can fudge a bit of resiliency by using storage pools. This would allow you to prevent a logical volume 'roaming' beyond a defined set of members. E.G storage pool 1 contains members 1 & 2, storage pool 2 contains 3 & 4, if 4 fails then volumes in pool 1 are unaffected. But that is a fudge.

You could use 3rd party software to replicate volumes to a DR setup... but when I looked at this the cost of the DR setup didnt' justify it. It was more cost effective to take the downtime hit and restore from backups.

But Dell are right, for the Member to fail in such a way it's un-recoverable is rare. Even if both controllers failed, provided the config is backed up they're FRUs and can just be swapped, config uploaded and back in business. The only scenario that would result in permanent data loss is simultanious multiple drive failure. To get that you'd be REALLY unlucky.

Can't say I agree with any of this - replication with equallogic units is pretty hot, it's included in the licensing and there's the HIT tools for VMware and microsoft allowing for DR of units. It's easy to configure and pretty solid.

Version 6 of the firmware introduced SyncRep which allows for volumes to be replicated to a second unit which if for critical volumes reduces the single SPOC issues.

Don't go above 90% usage or IO will drop significantly.

Dell will also provide you with loan kit to migrate volumes to so you can perform the firmware update with little concerns.

equallogics are very good and the service from dell is rather hot.

Edit - I don't work for Dell, but use EQL's a fair bit... I'd be more inclined to use pools to (hopefully) avoid the connections limit (1024).

 

[ecksmen]

Just that I've read, for Veeam to do SAN based backups, the veeam server needs `read only` access to the EQL volumes that have the VMFS on - But in the EQL grpadmin, you cant set access rights on an `per initiator` level, only at a volume level. I'm a little worried that the physical Veeam server has the potential to trash the vmfs disks. I'm probably being too risk adverse, but it would be good to hear of other peoples experiences.

ta

You send a snapshot to the backup server, not the live volume.

 

[ecksmen]

Don't forget to install San HQ!

 

[BruBoy]

We use EQL PS with Veeam and played around with the SAN backup on a windows box. In the end, we found it was simpler to just use the default veeam network backup which is fast enough for our environment (backing up to a linux VM with mounted NFS shares from our NAS boxes).

 

[brainchylde]

Veeam with EQ here - SAN based - very good. No problems. Reliable. I rarely get any backup failures.

My only problem with Veeam - and I know people say that tape is on the way out, but it is valid for us - is that it doesn't do integrated tape backup. We to D2D and then D2T which is a bit convoluted, but it works.

 

[topher]

replication on EQL does work using hit kit/asm the only thing to be aware of is the EQL's have a block size on 16mb, so if you want to hold lots of snapshots or replica they will take lots of space.. compared to something a netapp san which has 4k blocks!

If you are using SQL or Exchange the logs and DB's can be placed on iscsi volumes and snapped/replicated.. restoring or mounting a sql DB is nice and easy