Virtualization Megathread V2: VMs inside VMs

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

«‹›5 »

Erwin: Feb 17, 2006

The source machine needs to talk to ESXi on port 902. http://www.vmware.com/pdf/convsa_50_guide.pdf

See "TCP/IP and UDP Port Requirements for Conversion"

I guess it makes sense that it would hand off the big traffic to ESXi directly, so vCenter doesn't have to relay tons of data.

# ¿ Jun 13, 2012 18:29

Adbot: ADBOT LOVES YOU

# ¿ May 9, 2024 13:41

Erwin: Feb 17, 2006

MC Fruit Stripe posted:

Is there any real concern about virtualizing your domain controllers and going 100% virtualized? I've never done it but I'm kinda wondering if we could just put everything on ESXi - or would I regret that for some untold reason?

e: Hell, if you think this is a stupid question, realize that I only recently realized you can P2V a vCenter server into the same group of hosts that it manages - I thought it had to be on the outside looking in, a-doyyyy

As long as you have some non-AD account on the hosts so you can get AD back up and running without a chicken and egg scenario, you should be fine (as fine as running vCenter in the cluster it manages - I do it, but it can make for a fun puzzle at 3am and should be avoided if you're big enough). I happened to have a brand new physical server I wasn't using any more (1 socket, so not a host candidate) so I threw a DC on there. If I didn't have it, I wouldn't have bought hardware for a physical DC.

Also, make sure whichever DC is your top-level NTP server syncs to an outside NTP server, or your whole infrastructure will drift rather quickly.

# ¿ Jun 13, 2012 20:51

Erwin: Feb 17, 2006

LmaoTheKid posted:

Is Thinware vBackup any good for basic weekly backups of a few non critical VMs?

As above mentioned, my ragtag VM setup using local storage has been going along fine but it would be good to have a backup of everything.

I've got a poo poo ton of storage in another room on the other side of the office, so I can run backups once on Saturday when no one is using it.

Seems pretty decent for the free version, no?

I don't have any input regarding Thinware, but Veeam just released a free version of their backup - annoying as poo poo marketing campaign from them, but seems well-suited for one-off or once-in-a-while grabs: http://www.veeam.com/virtual-machine-backup-solution-free.html

# ¿ Jun 14, 2012 18:02

Erwin: Feb 17, 2006

Misogynist posted:

The biggest issue we had with Veeam was that, as many other people will tell you, CBT inexplicably fails to work in a huge number of deployments, leading to multi-day backup windows. Sometimes it works fine and then randomly stops. The net result is that your incrementals take just as long as your fulls. Ick.

They fixed the CBT issue a few patches ago, and I haven't had trouble with that since. Veeam has its issues, and I figure once we get to 75 VMs or so, I'll be looking elsewhere, but for my size, it's perfect, and it has some really cool features that I don't even need. The instant restore is pretty brilliant.

# ¿ Jun 15, 2012 13:47

Erwin: Feb 17, 2006

The office we use as a DR site is moving, and I'd like to take this opportunity to either colocate our equipment somewhere, or use some sort of hosted VMware infrastructure. Our needs are very small - physically it's 1 host and 1 disk array of about 5TB. Virtually it's 10 VMs, 7 of which are off all the time unless we fail over.

I know there are companies that have a VMware infrastructure to which we can connect and replicate to using Veeam, but for whatever reason I can't figure out the Google terms to use to find these people. I think vCloud director gives the ability to bill based on usage, meaning in our situation, our bills would be relatively low until a failover event. This does exist, right? Anyone have thoughts on the pros and cons of this vs. colocation? Is colocation definitely going to be cheaper?

# ¿ Jul 5, 2012 19:24

Erwin: Feb 17, 2006

Corvettefisher posted:

I know some people also poor man storage HA doing windows DFS=>NFS, which...works.

Just because something works doesn't mean you should do it. I wouldn't count this as a reason to use NFS. What an awful idea.

# ¿ Jul 9, 2012 21:22

Erwin: Feb 17, 2006

Looking to add a new VMware host. Our current hosts have Xeon X5660s at 2.8GHz. The Sandy Bridge Xeons available in Dell's 12th gen servers seem to overall have slower clock speeds than the Nehalems. It seems like virtualization is one place where clock speed still matters, and I should shoot for something at or over 2.8GHz, right?

# ¿ Jul 27, 2012 16:45

Erwin: Feb 17, 2006

Misogynist posted:

Are you bottlenecking on CPU speed? It's by far the least utilized resource in our environment. We average about 20-25% across the cluster.

Not at all, no. We're about the same, but my understanding was that if you have two identical processors, save for the speed (eg X5650 and X5660), that a VM with 1 vCPU will run faster (if it's maxing) on the faster processor, regardless of the free resources on the cluster. I know that CPU speed alone is not the determining factor with non-virtualized environments, but I wasn't sure if a single VM with a single vCPU would be negatively impacted by a slower clock speed on a newer processor.

We do have some Matlab build VMs that are CPU (vCPU) bound when working.

# ¿ Jul 27, 2012 20:07

Erwin: Feb 17, 2006

Why is VMware presales so god damned hard to get a hold of? Take my money already.

I'm reading up on VMware View and I'm trying to get an answer to the vagueness of the documentation. It seems to imply that, even if I have an existing vSphere infrastructure with a vCenter server, I need a separate vCenter server for the View VMs. Is that true?

It is even more vague about needing a separate set of ESXi hosts. I hope that's not the case, because that's dumb. I only want to virtualize 20 desktops, max.

The presales person I chatted with online had no idea, and calling presales just leads me to my local sales rep, who hasn't returned my call.

# ¿ Jul 31, 2012 16:11

Erwin: Feb 17, 2006

Corvettefisher posted:

My company is a reseller(or in the process) I would be happy to give you some priority.

What do you need to know I work a bit with view, we have a mostly Enterprise plus Enviroment and use the same vCenter standard to manage the view hosts as the EPP hosts. No problems thus far, you license them like normal. I believe as long as it is standard you are good

My questions were all in my post. Do you need a separate vCenter? Do you need separate hosts?

# ¿ Jul 31, 2012 16:39

Erwin: Feb 17, 2006

Pantology posted:

There are two ways to license vSphere for View.

The first is the standard vSphere licensing way. Your View VMs can be managed by the same vCenter server that manages the rest of your environment. Your View VMs can run on existing hosts, or you can purchase additional host licenses if you want to separate the workloads. Run all the VMs you can, but keep in mind that you're still subject to vRAM entitlements.

Thank you, this is exactly what I was looking for.

# ¿ Jul 31, 2012 18:20

Erwin: Feb 17, 2006

Shaocaholica posted:

Are any companies virtualizing workstations for computationally intensive tasks? Instead of giving everyone a 12 core machine that they will only tax maybe 20% of their workday.

Also, are there any thin client options that allow users to log into VMs without logging into a host OS(or dealing with any part of a host OS) first?

I'm beginning to look at this, too, but I don't have much information to share yet. Obviously you don't want to make 12-vCPU VMs. We're looking at some of Matlab's parallel computing options to offload computing to server hardware instead of desktops. Most of our workstation waste is in RAM, though, so it's perfect for virtualization.

As for thin clients, you probably want to look into zero clients. They run custom OSes instead of stripped-down Windows or Linux. I haven't demoed any yet, but don't they all do what you described? (Boot up -> log into VM).

# ¿ Aug 3, 2012 19:24

Erwin: Feb 17, 2006

Holy poo poo vCenter Web Client is a piece of poo poo. You need Flash Player to configure it, and you have to configure it on the server on which it's installed. Yes, VMware, let me just install a lovely insecure browser plugin on a server you assholes.

Nebulis01 posted:

Tangent here, not an HPC person so it's a quesiton.

Couldn't you just give the workstation/server a couple decent gpgpu cards? and be better served for heavy computational tasks?

I think you need to use a CUDA library in that case, and the licensing options are different. I don't know much about Matlab and I'm not doing the Matlab research, I'm just going to size hardware to the decision.

# ¿ Aug 3, 2012 19:28

Erwin: Feb 17, 2006

Shaocaholica posted:

Zero clients are pretty cool! Some questions though. Are there any that support more than 1 display? Higher than WUXGA resolution per display? What about portable clients (laptops) that are not just regular laptops with a dongle so you can maximize battery life and weight?

From what I've seen in my little bit of looking, Teradici has some that will do 2x 1920x1200, soon to be 4x, and 10zig has something that will do 4x 1920x1200, but there's some weird thing you have to do, like two linked sessions or something. I think I'll be asking for some demo units soon, and we definitely need a minimum of 4x 1920x1200 here to make VDI viable.

# ¿ Aug 3, 2012 23:41

Erwin: Feb 17, 2006

Docjowles posted:

I noticed today that some of our Linux guests have zero swap space provisioned within the OS. That's...not good, right? Or does the hypervisor completely take over swapping duties? Running under vSphere 4.0 (yeah I know) if it matters.

Hypervisor swapping is dumb swapping. It'll throw down to disk whatever it can grab, whether it's accessed a lot or not. The OS will be much smarter about what to swap, so you should definitely have it do so.

The hypervisor won't swap unless it needs to, and first it will try to use the balloon driver to force the OS to swap more intelligently. So also make sure that VMware tools is installed.

# ¿ Aug 9, 2012 21:29

Erwin: Feb 17, 2006

Nothing wrong with talking about Virtualbox. Can you ping the VM from the host OS? Can you ping the host from the VM? Can you ping 8.8.8.8 from the VM? If any of those answers are yes, it's less of a virtualization problem and more of a basic networking problem (probably). If the answers are specifically no, no, yes, then it's more of a virtualization problem (NAT instead of bridged connection, wrong DNS settings).

# ¿ Aug 22, 2012 22:07

Erwin: Feb 17, 2006

They've been marketing the poo poo out of the "Big Announcement" on Monday at VMWorld, and they want everyone to watch the live stream! I loving hate when companies do that poo poo. You're not Apple.

# ¿ Aug 23, 2012 14:41

Erwin: Feb 17, 2006

madsushi posted:

9) Graphics - The ability to leverage specific NVIDIA GPUs for 3D performance.

Is this for View? We recently did an initial test of View, and after 30 seconds of dragging a window around the screen, we threw up a little in our mouths and put it on the back burner.

As for shared-nothing vMotion, are there restrictions? Like, can I move a running VM to my DR site 500 miles away through a 35mb/s tunnel?

edit: VVV I'm not sure I can think of a use case for WAN vMotion, so meh v :shobon:

Erwin fucked around with this message at 01:36 on Aug 28, 2012

# ¿ Aug 28, 2012 01:09

Erwin: Feb 17, 2006

Does the web client still require a separate VM to serve it up like it does in 5? :allears:

# ¿ Aug 28, 2012 02:11

Erwin: Feb 17, 2006

FISHMANPET posted:

And if it ends up being that we need two identical servers then we can buy another server, I just need to know if that's necessary or not.

I just now finished putting in service a new Sandy Bridge host in my all-Nehalem cluster, so I figured I'd try this for shits and giggles. When I tried to migrate the secondary to the new host, I got

quote:

Host CPU is incompatible with the virtual machine's requirements
Mismatch detected for these features:
CPU model does not match

So looks like no? Caveat is that this is the first time I tried setting up FT (edit: in this environment), and there's no operating system (and thus no VMware Tools) on the VM. EVC mode is set to Nehalem.

Erwin fucked around with this message at 23:37 on Aug 30, 2012

# ¿ Aug 30, 2012 23:24

Erwin: Feb 17, 2006

madsushi posted:

Less about the drivers for me and more about the CIM providers so you can monitor things like hardware RAID status (individual drives).

I can see the individual drive statuses on all of my Dell hosts, and they're all the standard VMware image. Unless I'm not looking for the right thing?

# ¿ Sep 12, 2012 14:10

Erwin: Feb 17, 2006

Mierdaan posted:

How long should I spend troubleshooting terrible datastore latency when using Dell's R610 integrated broadcom NICs before I just replace them? Do these things work OK for anyone?

No issues here.

Corvettefisher posted:

Are they broadcom by chance?

Really?

# ¿ Sep 13, 2012 22:50

Erwin: Feb 17, 2006

No, he asked if anyone has issues specifically with Broadcom NICs and you asked if they were Broadcom NICs. You seem like you're always in a hurry to post post post regardless of whether it'll prove helpful. You live up to your title, for sure.

# ¿ Sep 13, 2012 22:59

Erwin: Feb 17, 2006

Do your iSCSI targets show you which IPs the sessions are coming from?

# ¿ Oct 17, 2012 19:25

Erwin: Feb 17, 2006

How much time skew do you guys see between VMs? I didn't really think much of it, but our programmers have reported seeing up to 10 seconds between different VMs, causing small issues like builds not triggering on continuous integration servers if you quickly check in code in rapid succession. This is a mostly Windows environment, and everything syncs to a DC which syncs itself with an external NTP server. The DC is virtual.

# ¿ Oct 22, 2012 18:51

Erwin: Feb 17, 2006

adorai posted:

We use group policy to specify windows server time sync should use an NTP server rather than DC.

Interesting. I have some old physical hardware and plenty of rack space, so maybe I'll just make a physical NTP server so I don't have to worry about it. And actually, I've got one physical DC, so maybe I'll make that the time server (currently it holds no FSMO roles and is really just there just in case).

# ¿ Oct 23, 2012 00:32

Erwin: Feb 17, 2006

Veeam released a new version of their Backup & Replication software today or recently, and I think their support is swamped. When I called earlier for an unrelated issue, someone had changed their hold music to circus music :laugh:

# ¿ Oct 26, 2012 17:31

Erwin: Feb 17, 2006

forge posted:

The reason I use the client is because with over 1000 pc's, servers 50+ switches routers etc. We are a two man shop. So I generally use the client the other guy I get to rdp.

You do know you can do two RDP sessions at once without RD CALs, right? Could your blank screen problem be as simple as the VMs being set to "turn off monitor" after 10 minutes or something in power settings?

# ¿ Oct 31, 2012 19:40

Erwin: Feb 17, 2006

forge posted:

I've done some more detective work here. So it seems like once it becomes unresponsive on a black console screen. The only way to wake it up is to go to something like say a legacy novell vm. Since these usually show up and don't have the vmtools installed. Go to the console on the machine I want to wake. Then go to the console on the legacy novell box. Then once it shows the console on that screen. I click the vm in question and it's already on the console screen. Then it changes the mouse cursor to a hand and allows me to click the screen and then it displays.

So it sounds like your VMs are not unresponsive, it's just that the console connection to them doesn't work (that was actually clear from all your posts, but I'm just confirming). In other words, you can still ping, RDP, and access services on them, right?

# ¿ Oct 31, 2012 20:47

Erwin: Feb 17, 2006

forge posted:

That's correct, You just can't see anything in the console as if it didn't connect. Even though everything in the VM is absolutely fine.

Because it's quicker to click on a machine and do what I need on it? Then going to run, typing mstsc, typing the ip, logging in, waiting a sec or two, doing what I need. Closing that opening another repeat? With the client I click the vm and do what I need. Click another do what I need. When I'm not at my computer I absolutely will rdp. However it's just easier, to each their own. I don't dislike rdp, and I don't dislike the client. You have to remember my work environment is probably a lot different from yours.

We all think you should change your behavior because it's weird, but I think you know that now.

I think "unresponsive" is the wrong word, and maybe that's why VMware support has been unhelpful. Nothing is unresponsive, it's just a communication problem between the client and the VM. What happens if you restart the client? What if you turn Windows Firewall off on the VM and/or vCenter and/or your machine?

# ¿ Oct 31, 2012 23:46

Erwin: Feb 17, 2006

Windows 7 makes mstsc better. All the RDP sessions are stacked in one taskbar icon, and you can hover over it for thumbnail previews. Right-click the icon for a list of your recent sessions v :shobon:

# ¿ Nov 1, 2012 14:42

Erwin: Feb 17, 2006

Are there any resources specifically on using vCloud Director for software testing environments that are Windows-based? Our test environment is 4 Windows VMs plus a dependancy on a spearate file server, and I assume if I were to make it into a deployable vApp that would get walled off from the production network, I'd need to include a domain controller.

I'm looking for some kind of step by step of someone who has done this. Eventually I'd like each of our developers to be able to fire up a clean test environment any time they want.

# ¿ Nov 16, 2012 18:40

Erwin: Feb 17, 2006

1000101 posted:

What specifically do they need from the external server? If they just need to copy files in then that should be fine. If they need to change things on the fileserver though that can complicate things.

Also yeah if you want them to join AD you'll probably want to toss a domain controller in there if your application actually needs it. One of my customer's actually don't need AD to do the software development so windows machines are just provisioned with local accounts and they're good with that.

I dunno if any resources specifically exist that are publicly available but I can certainly answer any questions you may have.

They do need to change files on the file server, so I was picturing a template with the structure of the subset of folders they need. All the service accounts are domain accounts, so the domain will be needed. I assume the DC in the vApp should not be allowed to contact the real DCs, right? But it needs to be peeled off of the real domain to start with? I don't actually want computers joining and unjoining the domain willy nilly all day.

How do the developers interface with the test environment? Since I assume it's walled off, opening a hole for RDP would be annoying (since you'd need to know IPs anyway, since DNS won't update). Or do you stick a Windows 7 client in there and they use the console interface?

# ¿ Nov 19, 2012 15:58

Erwin: Feb 17, 2006

I'm starting to get ready to upgrade to vSphere 5.1, and I think I'm going to replace my vCenter server (it was already upgraded from 4.1 to 5, and meh). The documentation says that SSO, Inventory, and vCenter can all run in separate VMs, or together, or whatever combination, but doesn't really provide guidance as to how to make that decision, other than "small" environments can use one VM and "larger" environments should split it up. Those terms are not quantified.

We're super-small (4 hosts) and my initial thought was to use one VM so that it's easier to protect, but we use datacenter licensing for Windows, so there's no actual extra cost to using multiple VMs. Anyone have thoughts?

# ¿ Nov 21, 2012 19:58

Erwin: Feb 17, 2006

Separate SQL. 85 VMs, but probably 1/3 of those are replica VMs from Veeam, or old powered off VMs.

# ¿ Nov 21, 2012 20:07

Erwin: Feb 17, 2006

VUM is already separate. Is there a quantitative reason behind what you said, or just a hunch? If it really comes down to "should be", then I might as well separate the roles.

# ¿ Nov 21, 2012 20:15

Erwin: Feb 17, 2006

Corvettefisher posted:

If I was deploying it on that small of an environment, I would have it all as one, if SSO is unavailible you can't administer vCenter anyways. So for a small environment(which vmware usually associates the term small environment 10 hosts 100VM's).
http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=2034918

Really for that environment unless you have SLA's not stated for HA and an HA event breaks those SLA's installing SSO/vCenter/Inventory Services, just ups the resource overhead on your cluster, the administrative overhead, Tshooting steps required, and backup resources.

I mean you can go with 2 SSO servers in HA and affinity to speparate the VM's on different hosts, then do heartbeated vCenter servers, and have complete protection; but it depends what your environment needs.

I just really don't see a benefit with what you have stated to install it all on separate VM's for that environment, unless you need a highly available environment that exceeds what VMware's HA will do.

have you looked into the vCenter Virtual appliance by chance?

Fair enough, sounds like all on one VM is the way to go. I have not considered the appliance since VDR and SRM replication both left a bad taste in my mouth (I try not to stray too far away from just ESXi + vCenter now unless it's 3rd-party software). I don't have any problems with running it on my own Windows VM that the appliance would solve (that I know of).

# ¿ Nov 26, 2012 15:23

Erwin: Feb 17, 2006

Ashex posted:

sdelete -c

There's your problem. You wrote 0s to all free space, and since you wrote something to all the space, the VMDK grew to accommodate. Like Corvettefisher said, svMotion or shrink should work, and since you defragged, it should shrink to as small as you can get it.

# ¿ Nov 27, 2012 21:33

Erwin: Feb 17, 2006

Is it just me, or is installing vCenter 5.1 a million times more complicated than previous versions? I had to restart installation like 5 times, each time after running up against some fiddly SQL setting that was set wrong. The best part is that most of the errors had nothing to do at all with the actual problem.

# ¿ Dec 3, 2012 20:48

Adbot: ADBOT LOVES YOU

# ¿ May 9, 2024 13:41

Erwin: Feb 17, 2006

Separate MSSQL 2008 R2. I did a simple install, but all problems were SQL related. I ran the SSO table spaces script, but decided I didn't want the SSO database called "RSA" because that's very non-descriptive. But, turns out you can't change the database name. Then, I forgot to make a 64-bit DSN for vCenter, and after doing so, it wouldn't recognize it until after I quit and reran the installer. THEN, it threw an error that the SQL user I gave didn't have the right permissions on the vCenter database, even though it did. Turns out it also needed to be dbo on the msdb database to create agent jobs, but the error didn't say that.

The main problem is insufficient error descriptions.

three posted:

The method to install SSO with an external database is ridiculously stupid. VMware should be ashamed they shipped it that way.

Yeah, it's more convoluted than some of the shittiest open source software I've dealt with.

edit2: Also bravo on the lovely new web client requiring a separate VM with 4 cores to serve it.

edit3: And bravo on pushing this god drat web client and not having an Update Manager plugin for it :wtc:

Erwin fucked around with this message at 23:13 on Dec 4, 2012

# ¿ Dec 3, 2012 21:30

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

«‹›5 »