|
Kachunkachunk posted:That's really good to know, actually! Thanks. No worries, I think it is a Symantec issue since by default it asks the vShield device to dump hashes and rehash files. I don't recall having this issue with other VDI A/V's
|
#
?
Sep 26, 2013 00:39
|
|
|
|
| # ? Apr 29, 2024 06:42 |
|
|
Tequila25 posted:Okay, now that everything's been purchased and delivered, I'm in the process of setting up my first production ESXi cluster cluster. IMO, setting up a production virtual environment is one of the funnest things you can do in Infrastructure. Have fun! Protip: After you've set it all up but before you have any critical VMs on it, spend a day or two unplugging cables and powering off devices to simulate outages. It's a good sanity check to simulate some disaster scenarios to make sure everything is configured and working like you expect. Then, do the same thing but with your boss watching. When he sees the magical mysical vMotion maintain connectivity to a vm after a component failure he'll have faith in you and your gear and be able to show it off to his bosses, making you and him look good.
|
|
#
?
Sep 26, 2013 00:51
|
|
|
Agrikk posted:When he sees the magical mysical vMotion maintain connectivity to a vm after a component failure he'll have faith in you and your gear and be able to show it off to his bosses, making you and him look good. If your host didn't go down, unless you have vMotion set to put the host in maintenance mode based on a custom action off an alarm, it's not going to do anything on component failure. If your host went down, vMotion doesn't move the VMs. HA kicks in and restarts them, but it doesn't maintain connectivity. three fucked around with this message at 02:49 on Sep 26, 2013 |
|
#
?
Sep 26, 2013 01:47
|
|
|
eh
Dilbert As FUCK fucked around with this message at 13:11 on Sep 26, 2013 |
|
#
?
Sep 26, 2013 04:14
|
|
|
Can someone help me understand VMware licensing? I'm interning at a company which had several unlabeled desktops just scattered around for international users to remote into, so I thought it would be a good opportunity for me to play around with virtual machines. I installed ESXi on a ordinary dell desktop, configured it with vsphere and converted some machines with VMware Converter, and they seem to be working very well. The licensing for ESXi seems to be free since the machine it's installed on has less than 32GB and only 1 physical processor. Can I set up more ESXi boxes or is there a limit to how many free licenses you can use. And do I need a license for converter/vsphere? I've gotten everything from the official site and haven't been asked to pay anything yet. At the moment the machines aren't being used for anything other than my own personal testing, but I want to figure out the licensing specifics before I recommend it as a solution to my boss.
|
|
#
?
Sep 26, 2013 15:00
|
|
|
There is a free version of ESXi, you have to register on their site and get a license code. 5.1 has the limitation of 32gb, but that was removed in 5.5 which just went GA on Monday. Purchasing licenses will give you additional features (central management, APIs, vMotion, SvMotion, HA, DRS). Here is a breakdown of their licensing/features. http://www.vmware.com/products/vsphere/compare.html
|
|
#
?
Sep 26, 2013 15:21
|
|
|
Agrikk posted:IMO, setting up a production virtual environment is one of the funnest things you can do in Infrastructure. Have fun! I plan on doing this with the switches, SAN disks, hosts, etc. three posted:If your host didn't go down, unless you have vMotion set to put the host in maintenance mode based on a custom action off an alarm, it's not going to do anything on component failure. But Fault Tolerant VMs will not lose connectivity if I understand this correctly, right? I made a diagram for how I have my connections laid out on my two switches, but now that HOST1 is up and running, I realize that it uses controller 0 on the SAN for just about everything and the SAN keeps complaining that my second LUN is not on the preferred path, because all the I/O is going through controller 0. This should fix itself when HOST 2 is up and running through SAN controller 1, but is that a bad idea? Any suggestions on if I need to change my connections and how to change them? Edit: Or can I change the ESX config on HOST1 to change paths to use both controllers?
|
|
#
?
Sep 26, 2013 15:38
|
|
|
Tequila25 posted:But Fault Tolerant VMs will not lose connectivity if I understand this correctly, right? Correct, but very few people use FT due to its requirements and limitations (only 1 vCPU).
|
|
#
?
Sep 26, 2013 18:05
|
|
|
Tequila25 posted:I made a diagram for how I have my connections laid out on my two switches, but now that HOST1 is up and running, I realize that it uses controller 0 on the SAN for just about everything and the SAN keeps complaining that my second LUN is not on the preferred path, because all the I/O is going through controller 0. This should fix itself when HOST 2 is up and running through SAN controller 1, but is that a bad idea? Any suggestions on if I need to change my connections and how to change them? This doesn't sound right. I'm having trouble following your spreadsheet, but for sure your IOs should be on the preferred path on every host. In a multipathed setup, each host would be able to talk to each controller through each switch, if that makes sense. e: I looked back and saw you're running a MD3200i. I don't know much about it, but it sounds like you're getting a warning saying you're sending IO through the controller that doesn't own the LUN. That's a misconfiguration in any situation except a path failure. KS fucked around with this message at 18:58 on Sep 26, 2013 |
|
#
?
Sep 26, 2013 18:51
|
|
|
Tequila25 posted:I plan on doing this with the switches, SAN disks, hosts, etc. Dell MD SANs want you to always use the primary raid controller for everything. In the event you yank that controller out or it has a real failure, it will fail over to the other one. However, you'll continue to see "preferred path" messages until you move all LUN traffic back to the other (primary) controller. They do not handle load balancing across the controllers.
|
|
#
?
Sep 26, 2013 20:00
|
|
|
three posted:Correct, but very few people use FT due to its requirements and limitations (only 1 vCPU). And apparently they want the FT logging split out onto its own set of NICs too. Two for VM traffic, two for management / vMotion, two for storage, and now two for FT logging? I mean sure, if you've got two 10GbE NICs you're probably fine to combine everything but way to make FT even less attractive.
|
|
#
?
Sep 26, 2013 20:31
|
|
|
Think about what it has to do. It has to mirror ram. Do you really want that sharing with anything else? Even if you put it on 10gig you will want net I/O control/QOS to help prevent it from effecting other functions. We use FT in some corner cases. The biggest frustration I have (aside from the lack of SMP support) is that we can't control where the secondary VM ends up in the cluster. Half our hosts are in another datacenter and I'd prefer the secondary to stay opposite from the primary. At least in 5.1, there is no way to do that.
|
|
#
?
Sep 27, 2013 02:47
|
|
|
So I have a problem with my ESX 4.1 host and I'm not sure how to solve it. I currently have the Service Console port on the first NIC, and I want to trunk that port so I need to change the VLAN from 0 to 147. I tried making another Service Console port and using that to access the host to change the primary SC port, but as soon as I disconnect the original SC, I lose all connection to the host. Is there any way to mess with the Service Console port without migrating everything off the machine and putting it into maintenance mode?
|
|
#
?
Sep 27, 2013 03:14
|
|
|
FISHMANPET posted:So I have a problem with my ESX 4.1 host and I'm not sure how to solve it. Edit: Or do you not have a dedicated management port? Vulture Culture fucked around with this message at 05:21 on Sep 27, 2013 |
|
#
?
Sep 27, 2013 04:45
|
|
|
It's currently dedicated, but I want to change it so vmkernel is on that port as well. I can't change it in VMware and then change the switchport, because it complains that it's in use. I can't change the switchport and then use the second SC port I added to change the first, because as soon as I disconnect the first the second goes dead as well. E: And these aren't really in a cluster yet. They are a cluster, but they were originally setup standalone and I'm trying to get them clustered. E2: I could also setup a vMotion network, migrate everything off the machine, and then nuke it and start from scratch. God I hate this cluster. FISHMANPET fucked around with this message at 07:44 on Sep 27, 2013 |
|
#
?
Sep 27, 2013 07:39
|
|
|
parid posted:Think about what it has to do. It has to mirror ram. Do you really want that sharing with anything else? Even if you put it on 10gig you will want net I/O control/QOS to help prevent it from effecting other functions. Yeah, I get that it's pretty technically crazy to pull off. I just wish they'd use some crazy memory compression technique or something. Also, perhaps allow us to cluster over infiniband? Wouldn't give us any geographic diversity but infiniband would have enough bandwidth to lockstep a serious server with multiple threads. I'm really curious what you have that needs uptime so bad you need that kind of cluster but is light enough to survive on a single thread.
|
|
#
?
Sep 27, 2013 16:45
|
|
|
vCenter 5.1 to 5.5 upgrade wiped all my users and roles out. I only had 3, so not a big deal, but still annoying. Had to go in with the .local admin and add everything back.
|
|
#
?
Sep 27, 2013 18:40
|
|
|
Does anyone have any good metrics for guessing VM stun times based on average I/O delta or is it really a guessing game? The reason I ask is that I recently discovered a colossal issue in the infrastructure which my company is currently supporting. They are running the NetApp plug-in for vCenter which has an "optimisation and migration" task which was scheduled to run weekly on Monday @ 22:00. For each VM in the cluster this task snapshots them, "analyses" the snapshot to determine whether the VMDK is aligned correctly and then removes the snapshot. We have over 200 VMs with 125-ish of those in full-production. I had the privilege of being there last Monday night when this task kicked off and god drat everything went to poo poo. We run several failover clusters of which the majority have virtualised nodes. These were dropping out all over the place, so much so that some cluster groups and clustered disks failed to come back in an online state (I think one cluster lost the quorum witness disk on both nodes). There were several other services affected including the Exchange DAG, Lync front-end pool and SAP router. So yeah, when I saw that poo poo go down I dived into vCenter and disabled the hell out of the NetApp plug-in. Still, as an after action thing I'd like to know whether there is any way to predict stun times of VMs (Based on average VMDK I/O). I never thought that VM stun times could have such an impact but obviously it is still something that needs to be taken into consideration. Pile Of Garbage fucked around with this message at 18:58 on Sep 27, 2013 |
|
#
?
Sep 27, 2013 18:56
|
|
|
EoRaptor posted:vCenter 5.1 to 5.5 upgrade wiped all my users and roles out. I only had 3, so not a big deal, but still annoying. I actually had a similar issue last week with my 5.1 upgrade, apparently you can't use local Active Directory accounts with single sign on and linked mode. And it also broke any VMware permissions based on local permissions, so all our users who were VMware admins from being local admins on the vCenter box broke.
|
|
#
?
Sep 27, 2013 23:24
|
|
|
Anyone care to hear me blabber about some design I did and jot me on poo poo such as design decisions? Want to do it live so, yeah..
|
|
#
?
Sep 28, 2013 04:03
|
|
|
Dilbert As gently caress posted:Anyone care to hear me blabber about some design I did and jot me on poo poo such as design decisions? Do you still have a blog? I'll read stuff but live is a different beast.
|
|
#
?
Sep 28, 2013 06:58
|
|
|
Moey posted:Do you still have a blog? Yeah, I just need to do something with it. My goal is to do some mock defense of my design, just for fun.
|
|
#
?
Sep 28, 2013 20:28
|
|
|
El_Matarife posted:And apparently they want the FT logging split out onto its own set of NICs too. Two for VM traffic, two for management / vMotion, two for storage, and now two for FT logging? I mean sure, if you've got two 10GbE NICs you're probably fine to combine everything but way to make FT even less attractive. Not only for FT, but I would split vMotion and management as well. I've seen instances where lots of vMotion traffic caused the management agents to stop responding and hosts to vanish from vCenter.
|
|
#
?
Sep 29, 2013 21:39
|
|
|
How old are your guys's VMware servers? Looking to replace 4 servers running 2 x Xeon E5335's (4 cores @ 2.0GHz) with 48GB of memory each. (I think they are 6 years old) Our VAR (MoreDirect) suggested this: quote:My recommendations for ESX would to run DL380G6 or G7 servers. I will put you in our monthly email blast for HP specials. Wouldn't it be completely stupid to not go with G8's and just get 2 servers, with current 8-core Xeons? Each of our servers runs around 30% CPU usage and 60% memory usage. And the 4th server is kept as a cold-spare (?), it doesn't even run.
|
|
#
?
Sep 30, 2013 20:19
|
|
|
Bob Morales posted:How old are your guys's VMware servers? No idea why he would advise such old models. Currently I am really fond of the Dell R620.
|
|
#
?
Sep 30, 2013 20:22
|
|
|
Moey posted:Currently I am really fond of the Dell R620. Boss is a moron and won't cross-shop Dell.
|
|
#
?
Sep 30, 2013 20:23
|
|
|
I have a chassis full of BL25p blades. Do I win the prize for "most anemic VM environment"? It can't even run 64bit guests  But I mean I did literally find it lying in the garbage so I can't complain TOO much.
|
|
#
?
Sep 30, 2013 20:27
|
|
|
We're in the process of replacing 10 DL380G6 servers with dual 5570 Xeons and 72GB of ram with 6 DL380G8 servers with dual E5-2690 Xeons and 384GB ram each. The G6's are about 3 1/2 years old right now I think. Maybe 4.
|
|
#
?
Sep 30, 2013 20:29
|
|
|
Yeah, don't get G6s or G7s, get Gen8s. I moved our oldest ESXi hosts running on PowerEdge 2950s to HP BL460c Gen8 blades w/256GB of memory earlier this year and couldn't be happier with them.
|
|
#
?
Sep 30, 2013 21:20
|
|
|
Martytoof posted:I have a chassis full of BL25p blades. Do I win the prize for "most anemic VM environment"? A joint venture we're involved in got a rack of BL360p G1 blades and wondered if it was economical to run VMware on them. They apparently like copying my setups and since I run ESXi Ent+ on DL360 Gen8s, they also needed to run it - on EOL'd hardware that was surplus from some other part of the company. (They also tried to copy my bandwidth - since I was only pushing 5 Mbit in production traffic on my sites, they ordered that much bandwidth for hosting and their office. Capped of course, whereas my production can burst to 100 and our office has well over 1 Gbit to the world)
|
|
#
?
Sep 30, 2013 21:43
|
|
|
Bob Morales posted:Wouldn't it be completely stupid to not go with G8's and just get 2 servers, with current 8-core Xeons? Each of our servers runs around 30% CPU usage and 60% memory usage. And the 4th server is kept as a cold-spare (?), it doesn't even run.
|
|
#
?
Oct 1, 2013 00:28
|
|
|
Bob Morales posted:How old are your guys's VMware servers? In the class I teach, Gen 6 - Gen 7 6100/6200 CPU's, and 32-48Gb ram, + 10 1gb/s nics, racking some impressive Gen8's Saturday. For customers, it varies too much to say. adorai posted:If I was the only person who voted on my team, I would go with the cheapest current gen dual socket highest ram density single proc 8 core server with dual power supplies that I could get. I think when we were looking last, it was around $2000 or so for HPs that fit the bill, plus $600 for a 10Gbe nic and another thousand for 96GB of RAM. We could add a second proc and double the RAM later if we wanted, and VMware provides redundancy provided you can tolerate short bouts of downtime while HA kicks in during hardware failure (which lets be honest, even the cheapest HP server is probably not going to fail). Pfft, obviously you should spend 5x times an HP for some quality UCS's. What do you think we live in some kind of magical world where you can just rip and replace servers? BTW has cisco come out with any other reasons to by the UCS's other than the unified fabric and UC Manager? Really haven't messed with them much since my new job. Dilbert As FUCK fucked around with this message at 02:57 on Oct 1, 2013 |
|
#
?
Oct 1, 2013 02:51
|
|
|
adorai posted:If I was the only person who voted on my team, I would go with the cheapest current gen dual socket highest ram density single proc 8 core server with dual power supplies that I could get. I think when we were looking last, it was around $2000 or so for HPs that fit the bill, plus $600 for a 10Gbe nic and another thousand for 96GB of RAM. We could add a second proc and double the RAM later if we wanted, and VMware provides redundancy provided you can tolerate short bouts of downtime while HA kicks in during hardware failure (which lets be honest, even the cheapest HP server is probably not going to fail). I did the same thing for our Hyper-V cluster in 2010 with Dell R410's. They've since had a second processor added and been upgraded from 32GB to 128 GB RAM. Clustered virtualization has made hardware upgrades SO nice.
|
|
#
?
Oct 1, 2013 04:50
|
|
|
Can we have a biggest-cluster dick waving contest? Here's our cluster at our current DC (All HP BL620c G7 blades): We inherited this cluster and 342 of the VMs from the previous service provider who decided to configure things ultra-conservatively. The hosts all support Nehalem generation EVC mode yet they manually set it to Merom. Also all of the existing 342 VMs have CPU/memory hot-plug disabled and E1000 NICs. Here's our cluster at the new DC (All HP BL660c G8 blades):  We've only just got all the storage hooked up to the new cluster (We've got two 10GbE dark-fibre links between the DCs) and will probably be migrating stuff over in the near future.
|
|
#
?
Oct 1, 2013 12:13
|
|
|
cheese-cube posted:Also all of the existing 342 VMs have CPU/memory hot-plug disabled
|
|
#
?
Oct 1, 2013 13:47
|
|
|
Misogynist posted:FYI, there is small but non-negligible CPU overhead involved in having hotplug enabled, so it may make sense of your applications are CPU-intensive or latency-sensitive. Probably not, though! Not that I don't believe you but do you have a link that expands on this? I'd like to know more and I couldn't find anything from searching on Google.
|
|
#
?
Oct 1, 2013 14:22
|
|
|
cheese-cube posted:Not that I don't believe you but do you have a link that expands on this? I'd like to know more and I couldn't find anything from searching on Google. It seems like Duncan Epping mentioned it here. Not sure what he's basing it on either, but most references to hotplug overhead seem to cite that post as a source.
|
|
#
?
Oct 1, 2013 15:29
|
|
|
cheese-cube posted:Can we have a biggest-cluster dick waving contest? Here's our cluster at our current DC (All HP BL620c G7 blades): I've seen bigger. Actually, I haven't. We all just run single systems and that is mostly what I see. It is very cool to see the number of vmotions you have done and stuff like that though. I didn't know that was tracked somewhere.
|
|
#
?
Oct 1, 2013 15:44
|
|
|
Dilbert As gently caress posted:BTW has cisco come out with any other reasons to by the UCS's other than the unified fabric and UC Manager? Really haven't messed with them much since my new job. No, Unified fabric, VICs, and statelessness, all tied together with UCSM, are still pretty much it. But what else are you looking for?
|
|
#
?
Oct 1, 2013 17:01
|
|
|
|
| # ? Apr 29, 2024 06:42 |
|
|
cheese-cube posted:Not that I don't believe you but do you have a link that expands on this? I'd like to know more and I couldn't find anything from searching on Google. Here is a good article from Chris Wahl on CPU Hot Plug effects: http://wahlnetwork.com/2013/08/21/cpu-hot-plug-effects-on-vsphere-vcpu-count/ There's some good chat in the comments of that blog post too.
|
|
#
?
Oct 1, 2013 17:04
|
|