|
evol262 posted:Without starting at startup, or without starting at all? When I create a VM, the last screen is either create and start or cancel. There's no button to just create and then do nothing.
|
# ? Mar 10, 2016 15:27 |
|
|
# ? Apr 26, 2024 15:16 |
|
DoublePLayer posted:Hopefully attempting to play a video game on a VM isn't taboo here. Both of those game should work in a VMware VM. The only place you will run into trouble is really modern games, because the performance will start to suck depending on the game.
|
# ? Mar 10, 2016 16:54 |
|
Uh so is there no ovirt-node ISO for oVirt 3.6? 3.6 is the "recommended" download on the oVirt page so I was hoping to go with that. edit: Oh wait, did 3.6 literally just drop? I'll go with a nightly node iso I guess. some kinda jackal fucked around with this message at 17:37 on Mar 10, 2016 |
# ? Mar 10, 2016 16:55 |
|
We've got a number of small sites with maybe 1-2 hypervisors that are connected back over site to site VPNs to a central vcenter. I was doing some network digging at one of those sites that was complaining about slow performance and noticed that for the last month, the second "top talker" behind the backup host replicating data off site was the local ESXi 5.5 box. It looks like it's pushing maybe 2 gigs of data a night back up to the vcenter box (TCP over port 902), and usually this is around say 9:30-9:45PM to maybe 12:30AM. This seems to be a daily recurring process. Anyone have any idea what this is? I was under the impressions statistics were gathered more frequently. CBT related maybe? Backups are done on site using Commvault and then replicated offsite. Normally wouldn't be hugely concerned about this, but for a number of remote sites the only decent connectivity options also have transfer limits. Maneki Neko fucked around with this message at 17:59 on Mar 10, 2016 |
# ? Mar 10, 2016 17:56 |
|
Host access to other hosts for migration and provisioning Authentication traffic for ESXi and remote console traffic (xinetd/vmware-authd) vSphere Client access to virtual machine consoles (UDP) Status update (heartbeat) connection from ESXi to vCenter Server https://pubs.vmware.com/vsphere-50/index.jsp?topic=%2Fcom.vmware.vsphere.security.doc_50%2FGUID-ECEA77F5-D38E-4339-9B06-FF9B78E94B68.html
|
# ? Mar 10, 2016 18:05 |
|
Do your hypervisor or vCenter logs show any user / service accounts logging into vcenter or the hypervisor at that time? Do they show anything at that time? What kind of throughput are we talking about, Kbits, Mbits?
|
# ? Mar 10, 2016 18:08 |
|
Potato Salad posted:Do your hypervisor or vCenter logs show any user / service accounts logging into vcenter or the hypervisor at that time? Do they show anything at that time? Not sure what kind of throughput we're talking, this particular site has Merakis and they tend to smooth things out reporting wise to an insane degree, but the amount of data transfer is in gigabytes, roughly 60+ gigs a month for a relatively small host with 3 vms on it. Generally speaking the timing corresponds to our backup windows, which is why I was wondering if it was CBT or something related there. I was going to dig through the logs more, but was just curious if anyone had any off hand thoughts.
|
# ? Mar 10, 2016 18:31 |
|
HPL posted:When I create a VM, the last screen is either create and start or cancel. There's no button to just create and then do nothing. You can "Customize configuration before install", then dump the XML with virsh, and define it with virsh (which won't start it). You can also open an RFE, but why create a VM you don't want to use?
|
# ? Mar 10, 2016 18:40 |
|
evol262 posted:You can "Customize configuration before install", then dump the XML with virsh, and define it with virsh (which won't start it). Virt-manager doesn't allow you to use OVMF in a nice manner. I have to edit the XML file to get it happening.
|
# ? Mar 10, 2016 20:18 |
|
Something is very wrong with whatever distro you're using.
|
# ? Mar 10, 2016 20:52 |
|
Potato Salad posted:Got windows 10? I tried getting Hyper-V to work, but it kept claiming the hypervisor feature is not running and I couldn't figure out where's the fuckup. Could it be my firewall settings? Somehow? DevNull posted:Both of those game should work in a VMware VM. The only place you will run into trouble is really modern games, because the performance will start to suck depending on the game. I guess I'll try VMware next. Thank you for the replies so far!
|
# ? Mar 11, 2016 08:35 |
|
DoublePLayer posted:I tried getting Hyper-V to work, but it kept claiming the hypervisor feature is not running and I couldn't figure out where's the fuckup. You need to enable virtualization in BIOS most likely.
|
# ? Mar 11, 2016 13:04 |
|
I should've mentioned I checked that already, as well as making the service active through Powershell.
|
# ? Mar 11, 2016 13:09 |
|
Is there a way to configure the threshold that an ESXi host will start breaking up large VM pages in to small ones for dedupe/compression? Out of box appears that it doesn't do it until the host hits 90%+ memory utilization, and since we run two small clusters of three hosts it turns in to a nightmare of balloon/vswap churn when I take one down for maintenance. I know I can configure non-important VMs to not use the MMU virtualization instructions which will make it only receive small pages and accomplish roughly the same thing, but I can't trust the people provisioning the VMs to do it correctly and controlling it on the hosts seems cleaner.
|
# ? Mar 11, 2016 21:33 |
|
BangersInMyKnickers posted:Is there a way to configure the threshold that an ESXi host will start breaking up large VM pages in to small ones for dedupe/compression? Out of box appears that it doesn't do it until the host hits 90%+ memory utilization, and since we run two small clusters of three hosts it turns in to a nightmare of balloon/vswap churn when I take one down for maintenance. I know I can configure non-important VMs to not use the MMU virtualization instructions which will make it only receive small pages and accomplish roughly the same thing, but I can't trust the people provisioning the VMs to do it correctly and controlling it on the hosts seems cleaner. Change the mem.AllocGuestLargePage advanced host setting from 1 to 0 to force small pages and thus more memory sharing.
|
# ? Mar 11, 2016 22:18 |
|
NippleFloss posted:Change the mem.AllocGuestLargePage advanced host setting from 1 to 0 to force small pages and thus more memory sharing. Yeah, but that will do it globally to all VMs. I have a few large ones that are memory intensive and doing that carries a performance hit that I would like to avoid. Lowering the memory pressure threshold would make the host go after VMs that are using their memory poorly while leaving the ones utilizing their memory well alone.
|
# ? Mar 11, 2016 23:15 |
|
I thought TPS was gone now.
|
# ? Mar 11, 2016 23:32 |
|
adorai posted:I thought TPS was gone now. Still there, just unlikely to get invoked.
|
# ? Mar 11, 2016 23:36 |
|
When setting up an oVirt node, how much swap should I be allocating, if any? Again, I'm coming from the ESXi paradigm where I don't think there is any. The host has 128GB of physical RAM and I don't think I'm going to hit the cap any time soon since this is a homelab. I can throw another 128 in if necessary. Zero should be OK, right?
|
# ? Mar 11, 2016 23:37 |
|
1000101 posted:Still there, just unlikely to get invoked. There's a "feature" for cloud providers where each VM salts its page hashes so they can't dedupe against each other out of box, justified as preventing some kind of data leakage condition where you could theoretically see what pages are shared between VMs and do something with that? It's a bunch of nonsense for most people doing self-hosting and you have to go in to the advanced host config to either configure a single salt for all the VMs you want to dedupe against each other or turn the salting off entirely. But yeah, TPS is broken out of box as of either 5 or 5.5.
|
# ? Mar 11, 2016 23:51 |
|
Martytoof posted:When setting up an oVirt node, how much swap should I be allocating, if any? Again, I'm coming from the ESXi paradigm where I don't think there is any. The host has 128GB of physical RAM and I don't think I'm going to hit the cap any time soon since this is a homelab. I can throw another 128 in if necessary. Zero should be OK, right? It's still Linux, so a little is safe. 2gb should be way more than enough
|
# ? Mar 12, 2016 00:22 |
|
1000101 posted:Still there, just unlikely to get invoked. There is a setting in Horizon View that allows reenabling it easily if desired in a VDI environment at least. Can even do it per pool, etc.
|
# ? Mar 12, 2016 01:30 |
|
DoublePLayer posted:I should've mentioned I checked that already, as well as making the service active through Powershell. If you run systeminfo, what are the last 4 lines of its output? All of them should say Yes.
|
# ? Mar 12, 2016 04:47 |
|
God dammit, GPU passthrough, why you gotta be so rude?
|
# ? Mar 12, 2016 09:38 |
|
So how the hell do I pass-though a USB controller? I just cannot get it working with the VM, all the USB devices plugged in on to the controller appear on the host and not on the client. It's a motherboard (Asus Maximus V GENE) built-in controller, ASMedia 1042 that has it's own IOMMU group. The GPU (960GTX) works just fine, I'm sure I'm missing something reaaally simple.
|
# ? Mar 12, 2016 19:31 |
|
e: I'm an idiot
|
# ? Mar 12, 2016 20:09 |
|
HPL posted:God dammit, GPU passthrough, why you gotta be so rude? On what platform? Kivi posted:So how the hell do I pass-though a USB controller? I just cannot get it working with the VM, all the USB devices plugged in on to the controller appear on the host and not on the client. It's a motherboard (Asus Maximus V GENE) built-in controller, ASMedia 1042 that has it's own IOMMU group. The GPU (960GTX) works just fine, I'm sure I'm missing something reaaally simple. Pass through USB devices individually unless you have a separate controller. IOMMU is not necessary for USB passthrough, and it's almost certain that the USB controller is on the same IOMMU group as a bunch of system stuff you can't pass through.
|
# ? Mar 12, 2016 20:27 |
|
Also, create an XHCI root hub and bind your USB devices to that. According to the QEMU guys, it's the most virtualization friendly in regards to overhead.
|
# ? Mar 12, 2016 20:47 |
|
evol262 posted:On what platform? Just about anything. So far I've been getting the closest with Manjaro. I'm having a problem with libvirt-sock not existing so the host can't connect to the hypervisor. Could be a polkit thing, who knows. I'm like 3/4 of the way there.
|
# ? Mar 12, 2016 21:52 |
|
If I have an esxi 4.1 setup, what's the quick and dirty on upgrading to the latest? Can I just shut down the vm's, plop in a cd\flash drive, upgrade, then boot the vm's right back up? I assume on first boot the vm's get updated and an esxi 4.1 can't boot them from that point, but is the process really that easy?
|
# ? Mar 12, 2016 21:52 |
|
HPL posted:Just about anything. So far I've been getting the closest with Manjaro. I'm having a problem with libvirt-sock not existing so the host can't connect to the hypervisor. Could be a polkit thing, who knows. I'm like 3/4 of the way there. So KVM, or Xen? This isn't really a passthrough problem, then. It sounds like manjaro is busted. Is libvirtd running?
|
# ? Mar 12, 2016 23:41 |
|
bobua posted:If I have an esxi 4.1 setup, what's the quick and dirty on upgrading to the latest? Can I just shut down the vm's, plop in a cd\flash drive, upgrade, then boot the vm's right back up? I assume on first boot the vm's get updated and an esxi 4.1 can't boot them from that point, but is the process really that easy? I don't know what the best practices are from that old of a host, but what I do is: 1) Shutdown all of the VMs 2) Remove them inventory 3) Install new ESXi on top of old one: do NOT upgrade. Replace. 4) Do the config on new host and then add the VMs back from the datastore 5) IMPORTANT: upgrade the virtual hardware before powering on. 6) Update VMware
|
# ? Mar 13, 2016 00:27 |
|
evol262 posted:So KVM, or Xen? KVM. As far as I can tell, libvirtd is busted too but I'm having a heck of a time getting it going. Remove and reinstall?
|
# ? Mar 13, 2016 01:28 |
|
Use a distro which does QE. Or at least take the libvirtd.conf from a known-working distro
|
# ? Mar 13, 2016 05:36 |
|
mayodreams posted:5) IMPORTANT: upgrade the virtual hardware before powering on. This isn't important, and you need to upgrade the vmware tools in the vm before upgrading the virtual hardware. Also I think that esxi 6 doesn't support vmfs3 at all, so if you want to keep your vms, you'll have to upgrade to 5.5, upgrade your datastore to vmfs5 then upgrade to 6. GrandMaster fucked around with this message at 06:42 on Mar 13, 2016 |
# ? Mar 13, 2016 06:40 |
|
evol262 posted:Pass through USB devices individually unless you have a separate controller. IOMMU is not necessary for USB passthrough, and it's almost certain that the USB controller is on the same IOMMU group as a bunch of system stuff you can't pass through. Combat Pretzel posted:Also, create an XHCI root hub and bind your USB devices to that. According to the QEMU guys, it's the most virtualization friendly in regards to overhead.
|
# ? Mar 13, 2016 12:35 |
|
Kivi posted:It's seperate PCIe USB3 controller, in it's own IOMMU group (#13) according to a iommu.sh script that I found. It should just work, right? It shows up in the VM device manager. You can check the PCI ID in: find /sys/kernel/iommu_groups/ -type 1 If it's on it's own group, pass it through. I assume you've done this. Does the USB controller appear in the VM and just USB devices plugged into it appear on the host? Kivi posted:Do you have guide handy for this? virt-manager can do this very easily, or I can give you XML if you want to edit the domain by hand. virt-manager can also handle passing through USB devices
|
# ? Mar 13, 2016 16:12 |
|
Kivi posted:Do you have guide handy for this?
|
# ? Mar 13, 2016 17:41 |
|
If you are passing as USB device to an older OS make sure the device supports that OS. Getting an Xbox controller to work with say Windows 98 doesn't sound as easy as an old serial Gravis.
|
# ? Mar 14, 2016 05:15 |
|
|
# ? Apr 26, 2024 15:16 |
|
BangersInMyKnickers posted:Is there a way to configure the threshold that an ESXi host will start breaking up large VM pages in to small ones for dedupe/compression? Out of box appears that it doesn't do it until the host hits 90%+ memory utilization, and since we run two small clusters of three hosts it turns in to a nightmare of balloon/vswap churn when I take one down for maintenance. I know I can configure non-important VMs to not use the MMU virtualization instructions which will make it only receive small pages and accomplish roughly the same thing, but I can't trust the people provisioning the VMs to do it correctly and controlling it on the hosts seems cleaner. I think it figured it out. mem.memminfreepct lets you manually configure the memory threshold before TPS and the balloon starts kicking in. The out of box default is a sliding scale where those features don't kick in until you're down to about 2gb free on a decent sized host. Fine for someone running a 20+ host cluster, but really stinks if you run a 3 host cluster with a lot of ram. http://www.vmwarebits.com/content/memmemminfreepct-calculation-example I'm going to set mine to 10% which will trigger the busting up and hashing of large pages at 70% host utilization which will hopefully make maintenance operations less painful.
|
# ? Mar 14, 2016 15:42 |