Storage Spaces Direct - Issues

[GreyFox]

Hi all,

 

We're currently looking at moving to Storage Spaces Direct, but I've been experiencing a few issues, mainly with speed. A bit of background first.

 

We've been having a lot of issues with our current clusters (one Hyper-V and the other for storage). I think the storage is ok - a DataOn setup with two Server 2012R2 machines and a huge storage pool. The Hyper-V cluster, with VMM (eugh!) consisting of three hosts, seems to get upset when it shifts VMs around on its own - such as when restarting for an update. They get stuck, can't turn on/off or move. We have to restart the hosts. Because of this we had been running off just the one host for a while. Not great, but they are beasts, so are capable, plus it was a lot more reliable than what we had some time ago! We're currently using two hosts, with one of them being a bit suspect. The third seems to be the one that causes the problems. This has had hardware changes, a complete rebuild etc, but seems to cause the majority of the issues. We've had an external support company working on this (for quite some time), who's been in contact with Microsoft. They say this could be down two one of things. One being something do do with our domain name and I can't remember the other. Either way, this is why we're looking to move.

 

To S2D. This is pretty new to me, so I've tried to follow a couple of guides. We've got two hosts, although three seems to be the most common way of setting this up. I've installed the correct adapter, and packed them with disks. We have a mixture of disks in each host, but they do mirror each other. In each host we have a couple of 300GB SSDs for the OS (mirrored), five 1.75TB SSDs and seventeen 1.2TB 10k HDDs. I've created the cluster and added a file share witness. Then created a storage pool. All looks good. For network, each host has two teams, 2x 10G for management, 2x 10G for storage.

 

I have managed to get it up and running nicely (ish), with a number of VMs moved over. Failover works nicely and I thought we were all good to go. However, the transfer speed is incredibly slow. Manually copying a file to the shared disk tops out at around 100MBps, but can be slower. Normal transfers to a non-cluster disk can be around 800MBps.

I’ve moved the VMs back and have started again. However, one thing I may have missed during the setup was the tiers. On the second attempt I noticed we only have a capacity mirror tier, with no mention of a performance tier, using the SSDs. They are showing up as ‘journal’ though, but I’m unsure if they are actually being used for anything. I don’t know if this is normal or not with just a two node setup? We have one or two very large servers to migrate across, or move, if they fail to live migrate. So writing at 40 – 100MBps is going to be an issue and may take days!

 

Speed and the missing tier are my issues. Any ideas please guys, or is this normal? I don't think we saw any performance issues when using the VMs. It seemed to be purely the migration/copying that was painfully slow.

 

Thank you

 

GreyFox

[PotNoodleTech]

Did you use Mirrored or Parity when you created the Storage Layout? Only reason I ask is I did a bunch of testing on 2012R2 storage spaces and the parity option gave good read performance but atrocious write performance on the SAS/SATA disk group. Went back to mirror and got brilliant performance again (though ob. lost 50% storage capacity).

[GreyFox]

Hi AButters,

 

Thank you for your reply. I believe I set this as Mirrored. We've only got half of the total storage space available, so that would support this. I don't think you could do Parity with just two hosts.

I can't remember what read speeds were showing, but the read-world write performance is as I mentioned above