|
It seems like something has happened recently to kick VMware into making non-poo poo ways of managing their products
|
#
?
Mar 28, 2016 21:51
|
|
|
|
| # ? May 9, 2024 00:09 |
|
|
Thanks Ants posted:It seems like something has happened recently to kick VMware into making non-poo poo ways of managing their products
|
|
#
?
Mar 28, 2016 21:58
|
|
|
Vulture Culture posted:Now if only they could make one for managing licenses and keys that actually works and is online more than 50% of the time If Cisco made a somewhat not-lovely SMARTnet portal, maybe there's hope for VMwawre.
|
|
#
?
Mar 29, 2016 01:28
|
|
|
Maybe this is super obvious but why does VMware use E1000 as a default NIC for W2012 VMs? Is it just because its detected OOB? I always install VMware Tools and it installs the VMxnet3 drivers anyway. Should I be using E1000 over VMxnet3 for some reason or is it really no big deal either way?
|
|
#
?
Mar 30, 2016 15:41
|
|
|
Martytoof posted:Maybe this is super obvious but why does VMware use E1000 as a default NIC for W2012 VMs? Is it just because its detected OOB? I always install VMware Tools and it installs the VMxnet3 drivers anyway. Should I be using E1000 over VMxnet3 for some reason or is it really no big deal either way? Unless something has changed, you should be using VMXNET3 for all Windows deployments. The E1000 adapter is an old emulated Intel gigabit NIC (which is why it works without VMware tools). VMXNET3 is a virtualized 10gbe NIC optimized for for VM performance.
|
|
#
?
Mar 30, 2016 15:55
|
|
|
Moey posted:Unless something has changed, you should be using VMXNET3 for all Windows deployments. The E1000 adapter is an old emulated Intel gigabit NIC (which is why it works without VMware tools). VMXNET3 is a virtualized 10gbe NIC optimized for for VM performance. Yeah that's what I do, it's just been a while since I deployed a Windows machine so I was slightly surprised that it was the default for a Windows 2012 VM as suggested by vSphere 6 
|
|
#
?
Mar 30, 2016 16:31
|
|
|
I think the defaults are aiming at the highest compatibility rather than the 'best' configuration.
|
|
#
?
Mar 30, 2016 16:33
|
|
|
Thanks Ants posted:I think the defaults are aiming at the highest compatibility rather than the 'best' configuration. This.
|
|
#
?
Mar 30, 2016 16:34
|
|
|
Historically, despite the better performance, there have also been a number of bugs that have only manifested in the paravirtual network adapters and not the emulated ones. e: This one was a doozy: http://kb.vmware.com/kb/2006277
|
|
#
?
Mar 30, 2016 16:52
|
|
|
The vast majority of my VMware environment is E1000 and most of the linux servers don't even have VMware Tools / open-vm-tools installed.  You should never use E1000, and I think VMware should make vmxnet3 standard going forward.
|
|
#
?
Mar 30, 2016 22:49
|
|
|
Vulture Culture posted:Historically, despite the better performance, there have also been a number of bugs that have only manifested in the paravirtual network adapters and not the emulated ones. We're having issues with virtual network adapters just randomly not passing traffic or only passing some type of traffic. Been on the horn with HP VMware, they've seen it happen, and their reaction is pretty much 
|
|
#
?
Mar 31, 2016 01:52
|
|
|
Well I'd love to read the article, but the VMware KB site is (yet again) down. Maybe that's the problem with the VMware KB, they run on vmxnet3 NICs 
|
|
#
?
Mar 31, 2016 03:47
|
|
|
https://imgur.com/OzGVKD2.jpg
|
|
#
?
Mar 31, 2016 04:19
|
|
|
Wicaeed posted:Well I'd love to read the article, but the VMware KB site is (yet again) down. It may actually just be that you've gotten the cookie of death from their website. I find that sometimes it'll tell me its down and clearing my cache or a different browser sorts it out.
|
|
#
?
Mar 31, 2016 04:21
|
|
|
Got bit by this one over the last 3 months, finally figured it out a few weeks ago and it hasn't happened since: https://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1028373 Any vm that got rebooted had a small chance of networking not coming up, ddn't matter what host, vnic type, os, etc. I finally found an event viewer entry that pointed me in the direction of a mac address conflict, we traced it down to a cisco switch on the other side of the building that had an ancient version of ios on it. Updated to the latest ios and it hasn't happened since. I really wanna murder the prior director/sysadmins that had a segmented network and brought it back to a flat layer 2 for our entire corporate office.  Not worth fixing right now, we're moving servers to a colo in the fall and potentially a new building in '17.
|
|
#
?
Mar 31, 2016 14:10
|
|
|
The audio over HDMI in my virtual machine crackles and pops, what's the best way to mitigate that? The DPC latency is quite high.
|
|
#
?
Mar 31, 2016 16:40
|
|
|
devmd01 posted:I really wanna murder the prior director/sysadmins that had a segmented network and brought it back to a flat layer 2 for our entire corporate office. Ok, this is hilarious. The layer 3 fad has passed.
|
|
#
?
Mar 31, 2016 17:06
|
|
|
Moey posted:Ok, this is hilarious. The layer 3 fad has passed. I mean, people spend so much time and money implementing layer 2 overlays, why not just tackle the problem at the source and get rid of layer 3 entirely!?
|
|
#
?
Apr 1, 2016 02:30
|
|
|
Less than 10 years ago we ran the entire network on a single /13 netmask. Worked great, lets go back to that.
|
|
#
?
Apr 1, 2016 21:25
|
|
|
BangersInMyKnickers posted:Less than 10 years ago we ran the entire network on a single /13 netmask. Worked great, lets go back to that. That is impressive. The largest broadcast domain in all the lands!
|
|
#
?
Apr 1, 2016 21:38
|
|
|
And everything used multicast! Because we only had hubs.
|
|
#
?
Apr 1, 2016 23:05
|
|
|
BangersInMyKnickers posted:Less than 10 years ago we ran the entire network on a single /13 netmask. Worked great, lets go back to that.
|
|
#
?
Apr 1, 2016 23:31
|
|
|
We had a dedicated router in each building on the network, but they were configured to do broadcast forwarding so everyone saw every single broadcast packet on the network barring filtered stuff like DHCP. Absolute mess.
|
|
#
?
Apr 2, 2016 05:27
|
|
|
BangersInMyKnickers posted:We had a dedicated router in each building on the network, but they were configured to do broadcast forwarding so everyone saw every single broadcast packet on the network barring filtered stuff like DHCP. Absolute mess. Okay, I'm a rank newbie and even that doesn't make a lick of sense to me.
|
|
#
?
Apr 2, 2016 06:13
|
|
|
BangersInMyKnickers posted:We had a dedicated router in each building on the network, but they were configured to do broadcast forwarding so everyone saw every single broadcast packet on the network barring filtered stuff like DHCP. Absolute mess. I'll bite. Why?
|
|
#
?
Apr 2, 2016 06:45
|
|
|
BangersInMyKnickers posted:We had a dedicated router in each building on the network, but they were configured to do broadcast forwarding so everyone saw every single broadcast packet on the network barring filtered stuff like DHCP. Absolute mess. Did you work at VMware?
|
|
#
?
Apr 2, 2016 06:51
|
|
|
KillHour posted:I'll bite. Why? Ghetto windows domain-without-dcs probably.
|
|
#
?
Apr 2, 2016 12:28
|
|
|
Moey posted:That is impressive. The largest broadcast domain in all the lands! My first real job was at a small rural wireless ISP. When I started there, the whole customer wireless network was a half-dozen /24s and everything was a bridge. I think our broadcast domain covered most of a thousand square miles.
|
|
#
?
Apr 2, 2016 13:35
|
|
|
Oh, someone plugged in a computer. I'm going to lunch while spanning-tree does it's thing
|
|
#
?
Apr 2, 2016 15:11
|
|
|
Bhodi posted:Oh, someone plugged in a computer. I'm going to lunch while spanning-tree does it's thing Like they would use any kind of loop prevention.
|
|
#
?
Apr 2, 2016 20:47
|
|
|
I've tried the whole day trying to get GPU pass-through to work on Linux Mint 17.3/Ubuntu 14.04 and I give up. I can't get the UEFI/OVMF stuff to work. I actually got the nvidia card detected properly in the VM but the versions of libvirt/qemu didn't support hidden kvm so I had to get something newer and I lost kind of track. I used libvirt 1.2.8, qemu 2.1.2 from here https://launchpad.net/~jacob/+archive/ubuntu/virtualisation?field.series_filter=trusty I tried compiling newer ones but I don't actually know what I am doing besides entering make so it had a few issues. On the upside, I figured out blacklisting nouveau solves a few issues in Linux Mint like not being able to add keyboard layouts for switching.
|
|
#
?
Apr 2, 2016 21:13
|
|
|
I've kind of given up on GPU pass through with my current equipment and started hoping virgl gets done sooner than later.
|
|
#
?
Apr 3, 2016 00:21
|
|
|
Riso posted:I've tried the whole day trying to get GPU pass-through to work on Linux Mint 17.3/Ubuntu 14.04 and I give up. I can't get the UEFI/OVMF stuff to work. What's wrong with ovmf and UEFI? Can we start there? Troubleshooting those problems is the best starting point, since stubbing out legacy is frankly a clusterfuck. For compiling from source, make && make install might be OK on Mint. I've never tried it. But there's probably also a ppa with newer packages you can use.
|
|
#
?
Apr 3, 2016 00:36
|
|
|
Riso posted:I've tried the whole day trying to get GPU pass-through to work on Linux Mint 17.3/Ubuntu 14.04 and I give up. I can't get the UEFI/OVMF stuff to work. I found specifying the qemu command-line directly easier in the long run than fiddling with libvirt. Libvirt is a layer of abstraction that's not necessarily needed. Did you try using the current version of the UEFI/OVMF BIOS? There's current RPMs here: https://www.kraxel.org/repos/jenkins/edk2/ It took me days to get everything running smoothly, but seems stable and performs nicely now.
|
|
#
?
Apr 3, 2016 00:45
|
|
|
evol262 posted:... stubbing out legacy is frankly a clusterfuck. Yeah this a million times, I could not get VGA passthrough working without using UEFI for the VM. It became apparent from looking at the kernel logs that the VGA arbiter stuff was not behaving well.
|
|
#
?
Apr 3, 2016 00:48
|
|
|
HPL posted:I've kind of given up on GPU pass through with my current equipment and started hoping virgl gets done sooner than later.
|
|
#
?
Apr 3, 2016 01:27
|
|
|
Combat Pretzel posted:VirGL in Windows? Hope you got mad patience. There's not a lot that I do in Windows that's GPU-dependent, so as long as it gets done within some reasonable timeline, I'm good.
|
|
#
?
Apr 3, 2016 01:43
|
|
|
cliffy posted:I found specifying the qemu command-line directly easier in the long run than fiddling with libvirt. Libvirt is a layer of abstraction that's not necessarily needed. Yes/no. libvirt isn't necessary (though it can make stuff like iscsi volumes a lot easier), and it basically just wraps qemu for local VMs. But once you have the qemu options you want, finding the appropriate libvirt stanzas and saying "start this VM on boot every time, give me easy snapshots, etc" is much easier in the long run than invoking qemu-kvm every time you reboot or writing your own service which is a bad version of libvirtd.
|
|
#
?
Apr 3, 2016 02:57
|
|
|
evol262 posted:What's wrong with ovmf and UEFI? The ppa is what I linked. In short, qemu rewards me with a black screen instead of the efi shell despite directly giving it the two ovmf fds in the cli. Virt-manager doesn't even think libvirt has uefi support, giving me ye olde yellow triangle on the option when I want to use it. I was using the pure ovmfs I fished out of the Ed2k rpm.
|
|
#
?
Apr 3, 2016 03:01
|
|
|
|
| # ? May 9, 2024 00:09 |
|
|
Riso posted:The ppa is what I linked. On mobile, I didn't check the link. Again, let's start from scratch. There's way too much I don't know about your environment, and you haven't given enough information. There's an ovmf package for Debian/Mint/Ubuntu. .deb distros apparently expect to find the ovmf firmware in /usr/share/ovmf, not /usr/share/edk2.git... Is that where you put the firmware? If so, virt-manager should see it, but I'll need to actually install Mint in a VM to verify that, and I won't have the image downloaded for about 20 minutes. For qemu on the console... Is vfio enabled? vfio_pci? pci_stub? When you get a black screen from UEFI, are you trying with -vga std or -vga qxl? Try to get it working without passing through a GPU first. Using -vga none with a passed-through GPU will get you a black screen every time until you actually get vfio working. Can you post your qemu cmd? E: I won't be testing Mint. Their installer just spins formatting a disk in qemu, which is ridiculous, and cements Mint's terrible reputation. I'm happy to try to help walk you through it over the forums, but when combined with their slow rebasing and security problems, I'd suggest that you may have better luck with Fedora, Ubuntu, Suse, Debian, or even Arch. evol262 fucked around with this message at 04:45 on Apr 3, 2016 |
|
#
?
Apr 3, 2016 03:40
|
|