Virtualization Megathread V2: VMs inside VMs

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

«‹›13 »

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

Moey posted:

This. Build a beefy workstation with some good parts and you will be set. You can virtualize ESXi installs as well to practice some HA/DRS/fun stuff.

In my home lab, I once accidentally Storage vMotioned a thin-provisioned OpenSolaris VM into an iSCSI volume exported by itself. Don't do that.

# ¿ Apr 10, 2012 03:03

Adbot: ADBOT LOVES YOU

# ¿ May 9, 2024 18:35

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

FISHMANPET posted:

I'm guessing the iSCSI volume was made on RDMs, and the part he Storage vMotioned was the vmdk holding the root partition/pool.

This one. I was able to pretty trivially get the disk back by booting from a LiveCD, though

# ¿ Apr 11, 2012 21:27

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

Timdogg posted:

I am also in the education arena ...wondering what kind of pricing you are getting for this Compellent hardware. Not looking for an exact quote, but I have no concept how much Compellent costs for education now that Dell bought them...so a round figure would really help me out.

Pricing is typically considered to be vendor-confidential privileged information, and it's rather damaging for vendor relations to share it. If you have a budget figure, call them up, and see if they have anything that fits your range.

# ¿ Apr 11, 2012 21:38

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

Timdogg posted:

Yeah, I hate getting sales people all excited because the last thing they ever give me is an actual price. I usually have to go through a sales presentation and end up pissing people off because they were completely out of our price range.

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.

# ¿ Apr 11, 2012 22:48

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

Erwin posted:

And when you give them a budget of $100k, that hardware they just sold to another guy for $75k will be shown to you for $95k because they "really sharpened their pencil and got you a great deal."

This certainly explains why I gave a vendor a budget of $1.5m for a recent project and got back a quote at $600k.

Vendor management takes people skills. If you're not able to competently negotiate, you will get taken advantage of, end of story.

# ¿ Apr 12, 2012 15:21

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

Kerpal posted:

Has anyone ever tried actually backing up ESXi? I'm a little late upgrading from ESXi 4.1 to 5.0 U1. VMWare's upgrade guide warns that you cannot rollback to a previous version once upgrading. Googling around people seem to suggest that there is no point to backing up ESXi. I know you can use the vSphere CLI script vicfg-cfgbackup.pl to backup the configuration. Would it then be possible to rollback by reinstalling ESXi 4.1 over 5.0 and using the saved configuration file? Our environment is very simple, a single host with 10 VMs, no vCenter, update manager, or plugins.

Are you on shared storage, or local disk? The ESXi footprint by itself will fit on a USB key, if you want to test the upgrade non-disruptively.

# ¿ Apr 12, 2012 17:11

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

Bitch Stewie posted:

I have to ask, why would you give any vendor your budget upfront?

My personal opinion is that if you have $20k to spend on widgets, and you tell a vendor that you have $20k to spend on widgets, they'll come up with a solution that costs $20k.

Personally I prefer to work on the solution with the obvious caveat that you both need to know that the solution is going to come back in the ballpark.

I had to laugh when we had our SAN. List is/was around $58k. Special bid came back at $43k. They sent us an evaluation unit that was supposed to be from an eval pool but they had none so sent a brand new sealed unit.

After three weeks we were told we could have it at $23k "to save the hassle of collecting it".

I wanted to tell them to gently caress off out of general principle, but we smiled sweetly, said yes please and bought it.

In my (biased) experience, I've found that the "never name a number first" angle tends to come from:

People who aren't comfortable negotiating better prices, and are just looking for the best deal up front with a minimum of human interaction
People who don't know the value of the product they're asking about so they shift their gaze whenever the budget question comes up
People who aren't even sure what they're buying

Vendors can smell #2 like blood in the water. If you've got a lovely predator of a vendor, there's a good chance that playing your cards too close to the chest will make the sharks more likely to give you a lovely price, not less. Generally, we maintain a very close relationship with our vendor contacts. We're very open about our projects and the problems we face, and we're typically able to negotiate very aggressive pricing because we cut the bullshit, we're direct, and we don't play games. If we're apprehensive about a new vendor, we might say something like, "We have a combined budget of $X between this project and this other project." The vendor should be smart enough to smell out the ballpark figure we're after.

From a professional ethics perspective, pricing is generally considered vendor confidential and it's damaging to vendor relations if you disclose the details of your solution to another vendor. However, there's nothing wrong with saying "Vendor X's nearline storage system came in at $420 per TB and we really need you to come up with something in the ballpark for us to consider your solutions."

Bottom line: you need to know what the product is actually worth, and you need to know the type of person you're dealing with on the sales side of things. If you have a suspicion that a vendor's product is going to fall outside of your budget, express that up front. Vendors appreciate when you don't waste their time as much as you appreciate when they don't waste yours.

Vulture Culture fucked around with this message at 22:27 on Apr 12, 2012

# ¿ Apr 12, 2012 22:21

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

Bitch Stewie posted:

I'm assuming the type/model of vNIC plays a part but I've never been 100% clear on the differences - AIUI Flexible and E1000 are 1gbps and VMXNET is 10gbps?

Ignore the speed reported to the OS, it's not relevant to anything. Any of the emulated vNICs (E1000, etc.) will perform as quickly as the server can emulate them, but there's higher overhead and traffic still has to go through the whole virtual networking stack. The paravirtualized vNICs communicate directly with the hypervisor, and in the case of VMXNET3, can ferry data between VMs at the speed of shared memory.

# ¿ Apr 17, 2012 21:40

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

Wonder_Bread posted:

I don't believe this is the case at all. It doesn't even make sense to me. The E1000 is an emulated 1Gb Intel NIC, I don't see how it could go faster.

"Top link speed reported by the interface" seems like a pretty dumb flow control mechanism to me, even TCP windowing is better.

# ¿ Apr 18, 2012 02:55

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

Sleepstupid posted:

Noob question: I installed ESXi(?) on a spare server and used VMware standalone converter to convert a physical laptop to a VM on the ESXi server. I want to make another "instance" of that VM, is that possible? Can I just copy the one I already converted or do I have to re-convert the same laptop again (which took over 3 days)?

You can do a cold clone through the command line, then add it to your inventory through the Datastore Browser. When you start it up, it will ask you if you moved or copied the VM files.

# ¿ Apr 19, 2012 15:25

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

Sleepstupid posted:

Any more info on this? I'm using the vShere Client to remotely connect to the ESXi server and I don't see any kind of "datastore browser".

You need to right-click your datastore (pick host, select Storage tab) and select Browse from the menu. I couldn't find it my first time using free ESXi either.

# ¿ Apr 19, 2012 16:48

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

Sleepstupid posted:

OK, I found the Datastore browser, copied the existing files to a new folder, created a new VM and pointed it at the new folder. Now when I try to start the new VM I get a blue-screen during windows boot. Did I miss anything?

Thanks for all the help

Pretty tough to say with absolutely zero information provided about what the bluescreen actually says

# ¿ Apr 19, 2012 18:33

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

HalloKitty posted:

Obviously the main bottleneck is RAM if you don't have enough. But once you have enough RAM, an SSD is definitely going to give it a huge boost.

Keep in mind the cardinal rule that a ton of people forget -- operating systems will use spare RAM as disk cache. If you're being overly thrifty and giving your OS instances just enough memory to run their applications without swapping, you're going to be paying for it in I/O.

# ¿ Apr 20, 2012 14:32

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

I'm more confused about why you wouldn't have just bought Essentials Plus in the first place and saved yourself thousands of dollars and gotten actual features. Who sold you this poo poo?

# ¿ Apr 27, 2012 19:13

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

Frozen-Solid posted:

I have no idea. All this licensing was done long before I ever started here, and I know nothing about the licensing and contracts beyond what the key told me after I pasted it into the server. I wasn't even the network admin until a few months ago. I'm just doing my best to make poo poo work and learning as I go

Get in touch with your sales rep and get this whole situation straightened out. This isn't over with; you're overpaying on your maintenance and support subscription every year and taking advantage of none of vSphere's licensed features until you get this resolved.

Going by standard pricing, you're probably paying more per year in maintenance for these two hosts than a new Essentials Plus bundle would cost you.

# ¿ Apr 27, 2012 19:19

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

Frozen-Solid posted:

Just talked to my boss. He's going to talk to his licensing guy, since our subscription renewal should be coming up soon anyways. He says we're not paying much per year for what we have, so he didn't sound too worried even if we've got more features than we can use. I guess we'll see.

If you're not running vCenter you're literally paying for features that come in free ESXi

# ¿ Apr 27, 2012 20:22

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

three posted:

I was under the impression that 10K SAS drives are pointless to buy. They provide 33% less performance, but are not 33% cheaper.

Both of these postulates depend entirely on whether you're talking about 2.5" or 3.5" drives.

# ¿ Apr 30, 2012 18:25

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

bull3964 posted:

One day, your price range for new hardware is in the $1k-$5k range. The next day, growth pushes you to the point where hardware is in the $10k-$50k range. Upper management has a hemorrhage because technology costs are supposed to be decreasing rather than increasing, etc.

Man-hours aren't free. Keep a log of how much time you're spending dealing with issues related to your lovely environment, and use it to prepare some materials on the ROI you expect to see with your new purchases. Consider the cost of downtime to the business in this report.

# ¿ Apr 30, 2012 19:57

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

Speaking of VDI, has anyone used the Windows 8 implementation of RemoteFX? How does it compare to PCoIP?

# ¿ May 4, 2012 21:51

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

Syano posted:

The hate against Veeam has come up in this thread recently and since we are on it again I have some new hate to throw its way as well. We recently had a problem where our nightly jobs would fail with some obscure error about CTP files. I sent in a support ticket and after two days of no response I posted my error to their community forums soliciting help only to have my post denied and being told I needed to read their 'rules' or whatever which apparently state you aren't support to post support items on their forums. Anyways, long story short: It took 2 WEEKS to get the problem solved. Thats 2 WEEKS of not knowing if my backups would be usable or not during failure.

I will tell you without any apprehension that this is par for the course. They're a completely marketing-driven company and care more about their PR than their customers.

Corvettefisher posted:

So what would you guys say the 'best overall' backup is?

We recently switched to PHD Virtual for our smallish environment (12 servers, ~180 VMs) and we've had remarkably few complaints so far. We abandoned Veeam as soon as the maintenance renewal came up at the end of our first year.

# ¿ May 7, 2012 14:24

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

Mierdaan posted:

Did they up the Fast Track course costs for 5.0? I'm looking at one right now and it's $5500 I thought it was more in the $3500 range?

Fast Track is an accelerated, extended-length course. The Install, Configure, Manage course is the one that typically runs closer to $3k.

# ¿ May 7, 2012 14:46

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

Kenshirou posted:

Ugh, vpxa has been such a bitch at work lately. I have to clear our all kinds of retarded snapshots and restart management so that it can reconnect once in a while. I'm sure it has to do with our SAN (Nimble) but I'm not sure exactly why since it only happens occasionally.

If vpxa ever hangs, it's usually because it's blocking on some kind of device discovery, typically with your storage. Check your logs on the ESXi host and make sure you're not seeing any weird hangs or device resets.

# ¿ May 7, 2012 14:54

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

Kageneko posted:

Anyone have any experience with Proxmox? I am getting some VMs tomorrow for testing with and I don't know what format they will be in. But they were built and used in Proxmox. I want to know if I can use them in VMware?

KVM VMs should go over without too much trouble, but OpenVZ "VMs" that don't have a kernel will be a headache to port over to a real hypervisor.

# ¿ May 7, 2012 22:38

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

adorai posted:

i have a qos question regarding UCS. We have 6x UCS rack servers each with 2x 10Gbe nics connected to seperate Nexus 5548 switches. We plan to carve these links up into vNICs and apply QOS to them. My plan is to have each port carved into 1x management, 1x vmkernel for NFS traffic, 1x VM traffic, and 1x vMotion. I was planning to give highest priority to NFS, second to VM, third to vMotion, and last to management. Does this seem reasonable?

I'd prioritize management over everything besides storage. I'd hate to have all of my VM administration fail because some VM got compromised and started spewing out UDP traffic at the rate of msblast.exe.

# ¿ May 8, 2012 04:00

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

Bob Morales posted:

Isn't there something in the EULA that doesn't allow people to re-sell VM's on ESXi?

Free (unlicensed) ESXi only. Plenty of places do managed services offerings on ESXi, including giant global companies like Savvis.

# ¿ May 8, 2012 20:30

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

Corvettefisher posted:

Anyone know any Good Xen server management apps for windows? Terminal is not these peoples friends

oVirt?

# ¿ May 9, 2012 05:41

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

FISHMANPET posted:

I can recommend you don't follow Scott Lowe on Twitter.

Scott Lowe was such a good blogger before he became an EMC advertisement

# ¿ May 18, 2012 03:32

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

Update: I'm pretty sure I'm receiving prank phone calls from Veeam support even though (because?) I'm no longer a customer.

# ¿ May 21, 2012 21:21

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

DrOgdenWernstrom posted:

vSphere standard is by CPU, so I believe you would need four licenses of vSphere standard to cover the cpu's, but it has a 32gb/license limit on RAM.

As for the vSphere Essentials Plus, if you're talking about the kit that includes vCenter. "VMware vSphere 5 Essentials Plus Kit for 3 hosts (Max 2 processors per host) and 192 GB vRAM entitlement" So 64gb/server

EDIT: VMWare noob here, so I might be wrong

He can have 192 GB in one server if that's all he wants to use, vRAM entitlements are cumulative and not averaged across your environment.

# ¿ Jun 6, 2012 22:47

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

Also remember that this is a vRAM entitlement, not a physical memory entitlement for the host. That means you have some room to keep free RAM for HA failover and whatever else you need it for.

# ¿ Jun 7, 2012 15:16

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

adorai posted:

Define cores to spare? If you have less vCPUs allocated than you have cores, then you will see no performance penalty. If, like the rest of the world, you are stacked up 5+ vCPUs to cores, you will see a penalty. A VM cannot run until it has all of the cores allocated to it available. So if you have a 2 vCPU VM that needs to run some work, and only 1 core is available, it will wait for the second core to become available. It will hold the idle core during this time. VMware's relaxed coscheduling can fudge this a bit, but for every clock cycle that one core runs, a second core must run as well.

When you read that you should allocate the minimum number of cores necessary, it's not just a suggestion. We size based on MHz. If a single core VM is running 50% CPU or higher for extended amounts of time, and the application is multithreaded, we will add a core. Otherwise it is just 1 core, no matter what. We regularly tell application vendors to gently caress off on their requirements.

Erm, not quite.

Relaxed co-scheduling only "fudges it a bit" if you're underutilizing most of your cores by, say, hammering a multi-CPU VM with a heavy-computing single-threaded workload. This causes stop-start-stop-start effects like shoe-shining an LTO tape. Most of the time, it is a really, really significant SMP performance boost over hypervisors that don't implement relaxed co-scheduling, precisely because it doesn't need to do what you're describing as long as the CPUs are reasonably equally loaded.

Vulture Culture fucked around with this message at 04:26 on Jun 8, 2012

# ¿ Jun 8, 2012 04:24

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

The very concept of virtual RAM per physical socket doesn't even make sense.

# ¿ Jun 13, 2012 04:19

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

LmaoTheKid posted:

So I just extended a 1 drive VM using gparted. Expanded the drive size, booted to the live CD, extended to the partition, rebooted, let the drive check itself and boom, all was well. I tried using diskpart but it wouldn't let me resize the system partition.

Now I need to do the opposite and shrink a drive. I cannot use VMware tools for this because it says it's disabled for the drive (these were P2Ved to a 4.1 machine because the hardware was dying) and then imported into 5.

My Virtualized setup is VERY limited, we're using local storage for about 3 machines which are not super critical.

so If I shrink the drive with gparted, can I just reduce the hard drive size in the Vsphere client and be good to go?

What guest OS are you running? Windows 7 and Server 2008 R2 can grow the boot partition through diskmgmt.msc.

# ¿ Jun 13, 2012 16:01

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

vty posted:

Veeam rocks in my testing, by the way.

How many VMs are you backing up?

# ¿ Jun 14, 2012 21:00

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

vty posted:

Not many, about 25.

At what numbers did you have issues? Speed?

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.

Veeam's actually a pretty neat product with a reasonable price level if you're able to get the support you need. (I've been keeping in touch with Michelle Randolph, one of their support managers, after she got word from support renewals that we went with a competitor's product. She's a total sweetheart and really does want to make things better. If you have issues with their support, try to get in touch with her directly.)

The real problem we had following our support issues was that trying to get any kind of real concurrency out of Veeam was an absolute maintenance nightmare. Veeam's architecture requires you to set up full-blown Windows Server VM instances in order to run the backup proxies, but the proxies can't dynamically take individual tasks off the queue -- they can only take complete backup jobs and cannot distribute VMs from those backup jobs. We weren't about to redesign our entire folder configuration in vCenter to accommodate Veeam's rear end-backwards way of parallelizing backups (we have ACLs tied in and everything), nor were we about to hardcode VM lists into our jobs and risk missing backups of new VMs. So, we switched to a competitor's product. PHD Virtual has a few similar limitations, but is at least a multithreaded architecture, so it requires 25% as much loving around. With 125 or so VMs being backed up, we haven't had to add a second virtual appliance yet.

It's not a bad product by any means, but it definitely has issues scaling up beyond mid-sized environments. Then there were the support issues I've posted about previously, which I forgive them for. (I'm still receiving prank calls from the "wazzaaaaaaaaaahhhh" guy to my office phone, and I'm pretty confident it's the Veeam tech who no longer works there as a result of my support case.)

PHD Virtual has its own issues -- single-file restores are really annoying, for instance -- but in terms of "is this poo poo going to not eat my backups and not waste hours of my backup engineer's time every week," PHD Virtual came out to be a clear winner.

Kachunkachunk posted:

Now, for this handful of select ESX/ESXi boxes, you probably want to turn off AD authentication and rely on local, too.

All our ESXi hosts are using local authentication. I have literally no conception of why Active Directory is a useful thing to have for ESXi when you should be doing 99.9% of your environment's management through vCenter. But part of that is my cynicism about how every single VMware feature will, through some freak incident, turn around and bite you in the rear end when you least expect it someday.

Vulture Culture fucked around with this message at 05:07 on Jun 15, 2012

# ¿ Jun 15, 2012 04:50

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

Corvettefisher posted:

Domain controllers are not reccomended to be virtualized. I can't say I have had problems with 2008 and later but 2k3 have problems.

Domain controllers are absolutely fine to be virtualized, Microsoft just has a very specific list of things you should not do with a DC (reverting to a snapshot being a very important one). These are typically related to USN incrementing and are fairly obvious once you stop and think about how AD replication actually works. Things were vastly different in the pre-VT days where you were really likely to end up in clock skew hell if you weren't really, really careful about the hardware you virtualized on because virtual handling of RTC interrupts was so bad.

Most of these issues are going away with Server 8 anyway. MS is even supporting snapshot rollback.

Vulture Culture fucked around with this message at 13:40 on Jun 17, 2012

# ¿ Jun 17, 2012 13:38

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

HalloKitty posted:

Edit: eh, followed back, it's for consolidating some servers that are running on single core P4s.. I don't see how that deal is all bad; of course I'd suggest a server with a warranty and so on, but I guess that's not an option here.

The deal's not bad, but consolidating an entire infrastructure onto a single point of failure with no warranty seems like a job-loss event to me.

# ¿ Jun 21, 2012 17:31

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

sanchez posted:

We have clients who would see that as out of the box thinking to save costs and be delighted.

They always do, at first, but unfortunately "I told you so" isn't a good future selling point for your services when you're dealing with morons.

# ¿ Jun 21, 2012 23:02

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

This definitely sounds like a backend storage issue, since most non-VM storage operations in ESXi are synchronous for some awful reason. Check your ESXi host logs on the impacted host, especially vmkernel.log, and see if you can correlate anything useful from your log events.

Check your dangling snapshots, as well -- I've seen issues where Veeam created a chain of snapshots on a VM about 30 deep and dragged the host's performance into the ground until we cleaned them up.

# ¿ Jun 25, 2012 20:16

Adbot: ADBOT LOVES YOU

# ¿ May 9, 2024 18:35

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

FISHMANPET posted:

I'd put in a vote for vSphere Client is complete poo poo at rendering anything.

It's not the client's fault per se; WPF uses hardware acceleration to render vector shapes like fonts and sometimes you end up with really weird-looking results.

# ¿ Jun 28, 2012 15:59

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

«‹›13 »