1

I have an iSCSI storage array that I “inherited” a couple of years ago when the person who managed it left my company. It’s a Celerra NX4, if you care to know. I’m not a storage specialist but I am learning (I have to, after all, and I enjoy it).

Since I knew nothing about iSCSI back then, I had to learn everything, but in that process, I probably made some mistakes in how things were set up.

The array has two storage pools: a bunch of 450 GB 15k RPM drives in RAID 5, and a bunch of 1 TB 7200 RPM drives in RAID 10.

The array is connected to the network via two 1 GbE connections per side, for a total of four.

The main usage for the array is to act as the storage backend for a small VMware cluster of three hosts.

The only security in use is a list of approved IQNs that can connect to each LUN.

So here are my questions:

I have a bunch of LUNs dedicated to various purposes:

  • Generic LUN to the 15k RPM pool
  • Generic LUN to the 7200 RPM pool
  • Dedicated LUN for the data drive of our file server which goes straight to the same 7200 RPM pool
  • Dedicated LUN for the data drive of a SQL instance which goes straight to the same 15k RPM pool
  • Etc.

Since every LUN ultimately goes to one or the other pool, was it dumb not to use one or the other “generic” LUN and just dispense with the dedicated ones?

Another question I have is whether there is any point to using separate iSCSI targets with separate IP addresses for various purposes (keeping in mind that we don’t do a lot of things that could be done in terms of security) ?

When things were set up initially, I took notes from how an old Citrix XenServer farm we had was set up, and each XenServer host had its own iSCSI target, so I did the same thing with the ESXi hosts, thinking “well, that makes sense, that way you probably get better performance” and left it at that.

Then I started thinking about it again recently since we’re looking for new storage, and it occurred to me that since all the traffic is using the same physical wires to the same physical NICs, creating more virtual IP addresses doesn’t really do anything except complicate things.

So should I just have one iSCSI target?

I appreciate you reading this and posting your thoughts :slight_smile:

@VMware

7 Spice ups
2

How many VMs, what are they running? Are you attached to the idea of a SAN? How much storage do you need?

3

I’m glad to answer your questions but I think you may have misunderstood: everything is already set up and running, and has been running for months at the very least. I was only posting how things were set up and if the way I did it is good, bad or neutral, generally speaking.

That said, we’re talking 3 ESXi 5.1 hosts with about 15 VMs doing various things, the most intensive of which is probably the file server serving ~200 people. There is also a physical SQL server instance which mostly uses DAS but uses about 500 GB of the 15k RPM pool over iSCSI.

Total the Celerra has about 6 TB – 3 TB in one pool, 3 TB in the other pool. And yes I am attached to the idea of a SAN since I can’t really get rid of it, but I’d be interested to know why you asked :slight_smile:

4

I was mainly asking because it sounds like you have many single points of failure.

5

Pair of switches, A/B runs, multiple hosts, two controller Cellera. Sure the Firmware itself, and the management plane/building are SPOF’s but I wouldn’t say “many” SPOF’s. Now you could argue that someone who’s not a SAN guy managing one is a nasty point of failure but thats another thing.

4 Spice ups
6

That’s what I need to do to get responses to my questions. Put something wildly incorrect in the question.

7

Actually, as far as points of failure go, I think I’m good. All the ESXi hosts have access to all LUNs, vMotion and svMotion work, pairs of physical connections are routed through separate physical switches, multipathing is enabled in VMware, and the Celerra’s architecture is inherently failure-resistant.

I’m confident about implementing the next array without introducing a single point of failure, but I still wonder about the best way to manage my LUNs, generally speaking. I understand all this stuff pretty well and I am a VCP, but that doesn’t make me a good storage guy, and I aim to be better.

2 Spice ups
8

If it works just leave it alone… There are just too many things to fix in the current setup.

9

Like what? The OP was hoping for some help with the “things to fix” part.

1 Spice up
10

I’m not planning on changing anything in the current setup, really. But I would like to know all the things you think are wrong with it, since we’re getting new storage in the next quarter and I’ll be the one setting a lot of it up.

11

Any further opinions?

12

I wish I could help more. My setup technically “works” but I was hoping to learn more from the responses you get.

13

Yep, likewise!

14

Still hoping for input :slight_smile: