Virtualization Megathread V2: VMs inside VMs

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > Virtualization Megathread V2: VMs inside VMs

«‹›8 »

evil_bunnY: Apr 2, 2003

nolicense ESXi 5 is limited to 32GB vRAM.

Corvettefisher posted:

Invest in 10Gb/E! It is pretty cheap now, your storage will love you!

Heh. Have you looked at the prices of decent 10GBE switches? I mean 10GBE is a good idea anyway, but it's not exactly commodity-priced yet.

Corvettefisher posted:

NFS - Network file share, one of the slower storage techniques but very cheap and easy to maintain, can be greatly improved by use of jumbo frames. TCP/IP based Speeds Vary on Network Speed, latency is a bit higher than others
iScsi - Block level access to data over the network, better performance than NFS by a long shot, gets a bump from jumbo frames but isn't as noticeable as NFS gains, TCP/IP based, Speeds from 10Mb/s-10Gb/s, latency is comparable to that of a normal disk if set up on a vlan, network, or subetnet made for storage. Cheap and easy to maintain, good if you have an existing ethernet network

Also, I'd be wary to generalize this. All the entry level applicances do iSCSI, but NFS is less of a management pain for vSphere. One of the problems of NFS is vmware not wanting to update their supported version.

evil_bunnY fucked around with this message at 15:03 on Feb 20, 2012

# ¿ Feb 20, 2012 11:56

Adbot: ADBOT LOVES YOU

# ¿ Apr 26, 2024 12:42

evil_bunnY: Apr 2, 2003

HalloKitty posted:

Edit: for some reason I thought 4's free entitlement was 192GB, actually, it is 256:

Isn't that for 2 CPUs?

# ¿ Feb 20, 2012 17:34

evil_bunnY: Apr 2, 2003

Corvettefisher posted:

Ah, it works differently than I though, I was under the impression it did a different type of install good to know this thanks! Really doesn't feel that way when I install hyper-V from the add roles, didn't realize it went that deep

Yeah it basically jumps from its chair and swiftly slides a virtual pillow under its butt before landing again.

# ¿ Feb 21, 2012 00:14

evil_bunnY: Apr 2, 2003

Something like this?

http://www.virtu-al.net/2009/06/22/powercli-snapreminder/

# ¿ Feb 23, 2012 18:48

evil_bunnY: Apr 2, 2003

Mierdaan posted:

Rob de Veij wrote a nice thing called RVTools that does a nice job of picking out all the things like old snapshots, out-of-date VMware Tools installs, etc etc, as well.

This deserves an OP spot.

# ¿ Feb 24, 2012 00:03

evil_bunnY: Apr 2, 2003

stubblyhead posted:

Any suggestions for pasting info from the clipboard into a VM console window in vSphere? I've been playing around with AutoHotkey, but I haven't been able to get anything working.

Install/update the tools on your VM and it should just work.

# ¿ Feb 24, 2012 13:52

evil_bunnY: Apr 2, 2003

Bitch Stewie posted:

How do you all do hardware monitoring of your HP and Dell (and others) vSphere hosts?

Dell has a vCenter plugin if that's the way you want to go.

# ¿ Feb 25, 2012 12:04

evil_bunnY: Apr 2, 2003

Bitch Stewie posted:

Which I assume simply means "fail the lot over and to hell with the consequences" so we just need to keep an eye on resources (which having just two hosts we obviously do anyway)?

Disable HA on the VMs you don't want to fail over. If it's too much for the one host, welp. Congratulations, you now get to fight for some budget.

# ¿ Feb 26, 2012 18:58

evil_bunnY: Apr 2, 2003

feld posted:

I once watched a friend try to find his crashed virtual center hiding somewhere on ~70 hosts. Was quite amusing. See, the OS kind of hung; it wasn't like the VM crashed completely and esx could restart it automatically....

You're really asking for it if you have 70 hosts and don't pin vCenter.

# ¿ Feb 29, 2012 12:49

evil_bunnY: Apr 2, 2003

Misogynist posted:

vCenter will only live on the cluster it lives on (hat tip, Yogi Berra)

"This is not the vCenter you're looking for"

The way I've done in the past really depended on the customer's situation.

Now that I'm in-house and planning their next implementation, it'll be virtual (asynch replicated to B&DR site).

evil_bunnY fucked around with this message at 15:34 on Feb 29, 2012

# ¿ Feb 29, 2012 15:26

evil_bunnY: Apr 2, 2003

cheese-cube posted:

IBM V7000

What's the price like on these?

# ¿ Feb 29, 2012 16:19

evil_bunnY: Apr 2, 2003

Alctel posted:

Ok, so this old chestnut, virtual, physical or appliance for VCenter?

More on the subject:
http://www.yellow-bricks.com/2012/02/08/distributed-vswitches-and-vcenter-outage-whats-the-deal/

# ¿ Mar 1, 2012 01:57

evil_bunnY: Apr 2, 2003

Good VM sizing post (originally for MSSQL, but a lot of it applies to everything you virtualize)

I find that the single most telling memory usage stat is page faults (what MS calls hard faults now). Once you get rid of (most of) those, watching application/service latency while you vary vRAM will give you a better idea of what works best.

Always start small.

# ¿ Mar 4, 2012 14:38

evil_bunnY: Apr 2, 2003

doomisland posted:

vSphere Server and whatever licensing for the hosts + features.

You mean vCenter.

doomisland posted:

Yeah, I can never seem to keep the names straight...

Don't feel bad, no one can.

evil_bunnY fucked around with this message at 01:15 on Mar 6, 2012

# ¿ Mar 6, 2012 00:50

evil_bunnY: Apr 2, 2003

luminalflux posted:

I went to one a few months back when they held the first one in Stockholm.

haha I was pondering going to that same event.

# ¿ Mar 9, 2012 02:48

evil_bunnY: Apr 2, 2003

FISHMANPET posted:

One step closer to real honest to god virtualization

I'm most probably getting some 720's in the coming weeks so I'll let you know.

# ¿ Mar 12, 2012 12:42

evil_bunnY: Apr 2, 2003

Corvettefisher posted:

Would you say it created business, customers, or revenue for you? Was it noticeable or subtle(i know this depends on area) ?

The only thing that convinced customers to actually give us jobs were reference implementations we'd done before. Not being a partner might cost you a project, but in my experience having the box checked won't generate business, just enable it.

We got pretty good at upgrading Hypah-hypah installs with Essentials kits or implementing/expanding/upgrading small scale virtualization project before we started putting our fingers in bigger pies. I think the trick for us was expand scale and technical depth in consecutive steps.

evil_bunnY fucked around with this message at 15:15 on Mar 12, 2012

# ¿ Mar 12, 2012 15:11

evil_bunnY: Apr 2, 2003

Misogynist posted:

Is there any reason you're not considering PHD Virtual? We're most likely switching over to them from Veeam in the next couple of weeks. (The stories I could tell you about Veeam.)

I meant to ask you: what happened that made you go from "rargh IBM die in the fires of hell" one year ago to considering buying their stuff again?

# ¿ Mar 12, 2012 21:04

evil_bunnY: Apr 2, 2003

Misogynist posted:

The IBM stuff that we were having problems with was the Midrange Storage series (DS4800, DS5100) which is rebranded LSI with some IBM tweaks and parts fulfillment.

Urgh LSI.

On another subject: I'm trying to put together some hosts, and hitting an information wall: can I run a 8GB and a 4GB DIMM per channel (for 96GB totalon a 2socket box), or a 8GB and a 16GB DIMM per channel (for 192GB total) and still run them at 1.6ghz? Dell website says yes, but my rep checked with an Intel presales guy and he said 1.3ghz max?

This is annoying because I'd rather not run all my hosts at 128GB (gently caress you VMware licensing).

# ¿ Mar 13, 2012 13:22

evil_bunnY: Apr 2, 2003

Start low and watch application performance is the general rule of thumb.

One question is why it needed to be P2V'ed instead of re-implemented as a VM.

Rhymenoserous posted:

It's not P2V'd, started fresh.

Ah. I'd start at 1 vCPU and vRAM according to there and the load you saw on the old machine. Without metrics it's kind of hard to tell you what to expect.

evil_bunnY fucked around with this message at 17:33 on Mar 13, 2012

# ¿ Mar 13, 2012 16:05

evil_bunnY: Apr 2, 2003

Bring a book?

# ¿ Mar 14, 2012 15:32

evil_bunnY: Apr 2, 2003

Corvettefisher posted:

Anyone else getting crashes with installing vmware tools on Windows Server 8 beta? I totally lost the GUI and it locks up shortly after boot.

http://www.yellow-bricks.com/2012/03/14/workstation-and-fusion-tech-previews/

# ¿ Mar 16, 2012 14:47

evil_bunnY: Apr 2, 2003

Beelzebubba9 posted:

So my question is how would this be effected by vMotion in case of a host failure? I assume ram states are persisted in case of the failover, and the worst case is there might be a brief loss of synchronicity between all of the parts of the database. Is there anything I should be worried about other than that?

With 1 link per host you'll never get to vmotion the ramdrive VM. vMotion works by scooting vRam contents from one host to the next.

Beelzebubba9 posted:

Obviously you'd be compensated.

This is usually how professional services work.

# ¿ Mar 16, 2012 22:56

evil_bunnY: Apr 2, 2003

Also if your users are crazy scientists who need 128GB VMs it helps to imbalance the RAM across hosts.

# ¿ Mar 22, 2012 18:13

evil_bunnY: Apr 2, 2003

FISHMANPET posted:

We won't ever be able to get researches to use VMs (partly because of how much NSF sucks) so it's going to core infrastructure only for us, as well as a whole pile of storage (because somehow my boss has tricked everyone so that we can sell storage to researchers).

None of the stuff we do is that IO intensive (of course once I've got everything set up they'll find something). But apparently for sequencing data analysis the more RAM the merrier.

# ¿ Mar 22, 2012 18:44

evil_bunnY: Apr 2, 2003

Misogynist posted:

IBM finally released their technical documentation on the x3550 M4 and this seems to be generally applicable to the Sandy Bridge platform:

Unfortunately, the LRDIMMs that actually exist on the market seem to overwhelmingly be 32 GB modules.

So either I'm blind or IBM doesn't sell these with 16GB 1.6GHz modules?

# ¿ Mar 23, 2012 11:05

evil_bunnY: Apr 2, 2003

Misogynist posted:

If you read between the lines, there are no 16GB dual-rank 1600 MHz parts on the market today that are rated to run at 1.35V.

OK? There's still plenty of 1,5V parts though. Doesn't IBM sell those?

# ¿ Mar 23, 2012 18:01

evil_bunnY: Apr 2, 2003

Why did you downgrade instead of using a free key?

# ¿ Mar 26, 2012 00:06

evil_bunnY: Apr 2, 2003

luminalflux posted:

How much are the disks used on an ESXi host if all my datastore is over iSCSI? Should I be overly concerned with disk performance?

Plenty of people run ESXi on SD cards.

# ¿ Mar 27, 2012 12:33

evil_bunnY: Apr 2, 2003

luminalflux posted:

Really? It doesn't swap to local disk at all?

You can swap to shared storage, and the VM's swap to their vDisks.

# ¿ Mar 27, 2012 13:32

evil_bunnY: Apr 2, 2003

Kachunkachunk posted:

Slight corrections, I had to go read up on it since it's fuzzy. ESXi host swapping happens on the configured scratch partition, if needed. I think it should pick one by default now, but you can configure it. This is also the only way you will have some persistent logging past reboots. It'll be to a datastore, basically.

VMs swap to .vswp files (default is their configuration file location, but you can pick an SSD or specific datastores). Finally, the Guest swaps to its own contained swap file/partition within one or more of its given virtual disks. The .vswp files are used pretty much last. I don't believe that host-side scratch swap is used at all for VMs, aside *maybe* from VM overhead (due to vCPUs, 32 vs 64-bit, etc) and stuff, but even then, probably not. I'm thinking it's just for heap/worlds/processes.

It's pretty good practice to install an extra vHDD per VM on a different datastore for guest OS swap files/partitions.

That way when you start replicating, you can specify a non-stupid schedule for the swap data instead of keeping it near-synchro like the data you actually give a poo poo aobut.

# ¿ Mar 27, 2012 14:53

evil_bunnY: Apr 2, 2003

The only downside to learning the VMWare side of things is that it'll make Hyper-V intolerable.

# ¿ Mar 28, 2012 16:12

evil_bunnY: Apr 2, 2003

I've said this before, but the first time your virtual environment saves your bacon/allows you to fix/upgrade hardware during business hours instead of when you could be boozing you'll stop thinking about the license costs.

# ¿ Mar 28, 2012 22:48

evil_bunnY: Apr 2, 2003

adorai posted:

We got two demo servers, and they seem to be performing well, except for a few applications that need single threaded performance.

Bulldozer single thread performance is horrible, but I wasn't expecting it to be this bad.

# ¿ Mar 31, 2012 09:19

evil_bunnY: Apr 2, 2003

No Safe Word posted:

I play with pretty high settings and it's there for me too and it's loving stupid as hell.

Build a whitebox with components know to work, dint buy a server.

# ¿ Apr 9, 2012 21:09

evil_bunnY: Apr 2, 2003

that SVM can never finish. The delta can never be reduced.

# ¿ Apr 11, 2012 19:59

evil_bunnY: Apr 2, 2003

Erwin posted:

Only because the way storage hardware (and most IT resources) is sold is bullshit.

All it means is it's going to take me 5 times as long to figure out what poo poo should cost. And what I pay for it will be public record (hello European acquisition procedures) so the whole thing is an exercise in futility anyway.

Hell in Sweden I can look at any public purchase made in the past, and if a supplier still sells that product they *have* to offer me the same deal.

# ¿ Apr 11, 2012 22:05

evil_bunnY: Apr 2, 2003

Misogynist posted:

That's why you give them a budget up front. I used to do this with recruiters all the time and it saved me a lot of heartache.

This is the thing I don't understand. In the real world, there are things I'm willing to pay extra for (at the cost of compromising somewhere else). Most people know what they need (if they did some sizing, which they should have) better than the sales "engineer" at the other end of the phone.

# ¿ Apr 12, 2012 10:00

evil_bunnY: Apr 2, 2003

Rhymenoserous posted:

Most people I know running ESX on a local raid are buying machines with SD card slots now so they can run the hyper visor off one of those.

Not just on local datastores :clint:

# ¿ Apr 12, 2012 20:46

Adbot: ADBOT LOVES YOU

# ¿ Apr 26, 2024 12:42

evil_bunnY: Apr 2, 2003

FISHMANPET posted:

I didn't realize there existed a cable with an SFP+ end. I assume that I can then plug those into an Intel SFP+ NIC.

Twinax. Yes you can.

FISHMANPET posted:

Though Dell doesn't currently offer an SFP daughter card on the 12G servers. Though it really doesn't matter, I guess none of this is going to happen for at least 2 years

Eh, you can get both Intel and Broadcom sfp+ pcie NICs on 620/720's.

# ¿ Apr 13, 2012 23:48

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > Virtualization Megathread V2: VMs inside VMs

«‹›8 »