|
Dilbert As gently caress posted:Ehh vMotion can chew up some poo poo when you throw it into maintenance mode, especially if it's running view Yea, but as you said, you can control that with NIOC so have methods of keeping vMotion from eating up all of your bandwidth. A single vMotion only requires 250 Mbs, or about 1/40th the bandwidth of a 10GbE link, and even then it only requires it for a very brief period. You can also set the number of concurrent vmotions such that they won't come anywhere near saturating the link. If your hosts take somewhat longer to go into maintenance mode it isn't really the end of the world. Especially given that even limiting vMotion traffic to a 1/4th a 10GbE link and limiting the rest for storage will still provide more bandwidth than a 1GbE link. We have about 600 ESX hosts and all of them run everything other than ILO over dual 10GbE links to Nexus 7k and 5k switches in a vPC. Never had any issues with link saturation.
|
#
?
Apr 25, 2014 22:32
|
|
|
|
| # ? Apr 28, 2024 22:05 |
|
|
Dilbert As gently caress posted:Running about 6 hosts 3 with 12 and 3 with 8 1Gb/s nics in the servers, two 3750's, and a SAN of Netapp, Dell, and soon EMC VNX 5400. Don't see a network topology but I'll assume the ESXi servers and storage are plugged into the same physical switch? People shouldn't be messing about with your training infrastructure and I'm not sure what "VDS becomes corrupt" means. Just keep your management interfaces on a standard vSwitch and you should be otherwise set. That said it sounds like you're throwing a lot of engineering at this problem with potentially questionable benefit. For example on NFS you're probably not going to see more than 1 gig of throughput on any given ESXi host unless you do some trickery using multiple subnets and multiple volumes and evenly distribute your virtual machines over those volumes. NetApp TR-3749 I think covers this implementation in some detail should you choose to go down that route. It's a netapp paper but the theory applies universally using any NFS system. Are you actually driving that much traffic to the storage? What are you measuring with? What determines if a workload is running on iSCSI vs NFS?
|
|
#
?
Apr 25, 2014 23:15
|
|
|
skipdogg posted:Anything I can check to make sure memory performance is the best it can be on our VMWare hosts? Finance has some new servers running a memory intensive OLAP server (TM1) and performance isn't where they want to be. Now starts the fight between infrastructure(my team), applications, and vendor on why performance is not up to snuff. Application and vendor don't understand VMware so they blame us. I'm 95% sure our VMware environment is setup optimally. ESXi 5.1u2 running on DL360p G8 w/ dual E5-2690's and 384GB RAM. Their main VM is running 4vCPU with 64GB RAM and none of the performance counters anywhere come close to showing a heavy load.
|
|
#
?
Apr 26, 2014 00:48
|
|
|
evol262 posted:You spin a new image which includes the VIB that has the driver for your card and use that for install, generally. This has been a pretty intense experience as a person with minimal Linux experience. I wasn't able to get it to work so far but I did have an exciting time mounting an iso, adding a .tar, changing the boot.ini and then turning it back into an iso. If I can't get it to work I'll just throw in a NIC that's on the HCL. Thanks for pointing me in the right direction. If I get it to work I might try to repeat the process using the powerCLI too.
|
|
#
?
Apr 26, 2014 03:59
|
|
|
http://www.v-front.de/p/esxi-customizer.html is what you want 
|
|
#
?
Apr 26, 2014 04:12
|
|
|
1000101 posted:Don't see a network topology but I'll assume the ESXi servers and storage are plugged into the same physical switch? I'm going to do a GUG just for you on the design I have. Don't think this counts us even; I still owe you some drinks and food! Dr. Arbitrary posted:This has been a pretty intense experience as a person with minimal Linux experience. Does anyone want me to make a ESXi White box image? I'd be willing to package an ESXi image for whitebox labs for the sake of a gently caress you to some people; I don't care how long it takes.
|
|
#
?
Apr 26, 2014 04:15
|
|
|
Martytoof posted:http://www.v-front.de/p/esxi-customizer.html is what you want Thanks. This simplified my efforts. That site also led me to the right tar to use. Now that I have an image that has tested as working successfully, I'm probably going to try and make it work manually.
|
|
#
?
Apr 26, 2014 06:02
|
|
|
adorai posted:If you are confident that it's not your SAN/Network/Hosts then stand firm. In my experience, 95% of the time the application runs like poo poo it's because of a poorly written query. On more than one occasion it has been a query in a stored procedure that the application vendor knew was lovely, normally updates, and forgot to do on our install. I want to 2nd this. If they say something is wrong with your infrastructure ask them what exactly they think is wrong and you'll be happy to look at it. If you want to up the ante ask to see their queries. If it's a commercial DB like Oracle there are TONS of build in tools for query and performance analysis. I worked at a Big Corp where a mess of lovely perl scripts and poorly written SQL queries brought an AIX box with 32 CPUs, 128g of RAM, and 8x 4g HBAs to it's knees. When we tried to call them out on it and asked for read-only access to the Oracle performance tools they wouldn't give it to us. I worked at another Big Corp where people didn't know that multi-threading was something you program at the application level and tried to blame the OS guys for "not allowing the application to run on more than 1 CPU." Also remember that with regard to hardware the application vendor has no skin in the game. It's easy for them to say "yeah we need 32g of RAM" because they don't know how to size their own app and they aren't the ones who are paying for the hardware but it will cost them manhours to fix their poo poo.
|
|
#
?
Apr 27, 2014 14:15
|
|
|
Yo I know this is VMware central but are there any resources for a developer looking to get familiarity with hyperv? I ask because work gives me free licenses of everything and I might as well do something with them.
|
|
#
?
Apr 27, 2014 18:23
|
|
|
Any word if we are ever getting MSSQL support for the VCSA?
|
|
#
?
Apr 27, 2014 23:11
|
|
|
Dilbert As gently caress posted:Any word if we are ever getting MSSQL support for the VCSA? Last I saw, it'll come as soon as Microsoft releases a supported ODBC driver for Linux.
|
|
#
?
Apr 28, 2014 02:55
|
|
|
Pantology posted:Last I saw, it'll come as soon as Microsoft releases a supported ODBC driver for Linux. But they have it for the decommed vCloud Director... I think vmware's just trying to stout "HEY WE DON'T NEED MS!!!" Also 3kVM/128 host/40 site upgrade this week from vmware 5.1 to 5.5 u1 Dilbert As FUCK fucked around with this message at 03:13 on Apr 28, 2014 |
|
#
?
Apr 28, 2014 03:03
|
|
|
Dilbert As gently caress posted:But they have it for the decommed vButt Director... It's almost like those are different products, run by different teams, doing things in different ways.
|
|
#
?
Apr 28, 2014 03:17
|
|
|
Pantology posted:It's almost like those are different products, run by different teams, doing things in different ways. *chough* Gee it's so weird.... If only there was some way to beta test.....
|
|
#
?
Apr 28, 2014 03:23
|
|
|
Pantology posted:Last I saw, it'll come as soon as Microsoft releases a supported ODBC driver for Linux. They've provided one for a long rear end time. Is it not officially supported? That wouldn't surprise me I guess.
|
|
#
?
Apr 28, 2014 05:56
|
|
|
You can a actually load that into the VCSA, support is pissed if you do it though.
|
|
#
?
Apr 28, 2014 06:06
|
|
|
Docjowles posted:They've provided one for a long rear end time. Is it not officially supported? That wouldn't surprise me I guess. Was pulling that from this. I still parse "Linux" as "Here be dragons," so I'm not sure why they can't make the supported Red Hat client work nice with SLES.
|
|
#
?
Apr 28, 2014 12:50
|
|
|
Pantology posted:Was pulling that from this. I still parse "Linux" as "Here be dragons," so I'm not sure why they can't make the supported Red Hat client work nice with SLES. Different library versions. OSX and Windows both generally ship with every library they need (hence winsxs, osx .apps being directory trees with libs). It's possible to do this on Linux, shove all your poo poo into some directory and use that, but it's considered bad practice. So if RHEL has libfoo.1.2.3 and SLES has libfoo.1.2.4, you need to recompile (or at least relink) against the new version. Vendors don't like doing this, which is why they (IBM, Oracle, etc) generally just support RHEL and SLES (Ubuntu, fedora, Debian, and others move too fast or don't guarantee ABIs/APIs will stay the same for a given major release). Still not sure why VMware won't approve it. It's probably "we haven't tested this, and the license on the driver prevents us from shipping it, so it might work, but you're on your own in an unsupported configuration if you do it"
|
|
#
?
Apr 28, 2014 13:55
|
|
|
And people accuse Windows of DLL hell. They haven't ever seen Linux library shenanigans.
|
|
#
?
Apr 28, 2014 14:01
|
|
|
Riso posted:And people accuse Windows of DLL hell. They haven't ever seen Linux library shenanigans. Linux doesn't actually have library shenanigans in that sense. "DLL hell" wasn't an empty accusation. Early versions of NT (and all of 9x) shoved everything in system32, tried to save memory by loading the same DLL, and assemblies didn't specify versions, so when Windows updated foo.dll or some application installed a different version of bar.dll, it potentially broke everything which linked against it. Build systems are good about saying "libfoo updated to a new version which breaks compatibility? Rebuild everything that touches it". RHEL and SLES only break that compatibility on major versions. If you're seeing " library shenanigans ", you may be doing it wrong.
|
|
#
?
Apr 28, 2014 14:24
|
|
|
evol262 posted:Build systems are good about saying "libfoo updated to a new version which breaks compatibility? Rebuild everything that touches it". RHEL and SLES only break that compatibility on major versions. If you're seeing " library shenanigans ", you may be doing it wrong. Except RHEL 6.5 which ships with backward-incompatible libs for OpenSSL in direct contradiction of Red Hat's policy as far as I can tell  Most of our boxes are still on 6.4 because of ~reasons~ and it's getting increasingly harder to find packages that will install on them because people have been rebuilding them against the newer libssl/libcrypto.On the bright side, CentOS 6.4 wasn't vulnerable to Heartbleed so that made that week suck a lot less for me.
|
|
#
?
Apr 28, 2014 16:57
|
|
|
Can I get a hint on VMWare Horizon? I'm trying to set up a basic proof of concept lab environment that'll eventually get recreated on real servers. I've got a host with about 5 VMs now. DNS/DHCP Domain Controller SQL 2008 server V-Center Server 2008 that's going to be my view connection server. Am I headed in the right direction?
|
|
#
?
Apr 29, 2014 00:33
|
|
|
Dr. Arbitrary posted:Can I get a hint on VMWare Horizon? Yeah that's the right direction. If you need any help let me know(I've done a poo poo load of view). You'll want a security server if you plan to do public facing.
|
|
#
?
Apr 29, 2014 00:55
|
|
|
All the VM stuff I've been doing at work has gotten my coworkers interested in learning. I can set up hardware but does anyone know where I can find some structured lessons? I'm probably going to find my lab manual and then modify it as needed but I'd love other options.
|
|
#
?
Apr 29, 2014 02:36
|
|
|
http://www.amazon.com/VMware-View-B...rds=vmware+view This book is pretty good. Probably can hunt down the ISBN that the vmware view class uses if you want.
|
|
#
?
Apr 29, 2014 03:03
|
|
|
Dilbert As gently caress posted:http://www.amazon.com/VMware-View-B...rds=vmware+view Oh, I might pick that up for myself but for my coworkers I'm looking for some structured guides to installing esxi to provisioning a vm and so on. I can probably write up everything myself and it'll be time well spent but I figured that if there's already good stuff out there, I'd save the time.
|
|
#
?
Apr 29, 2014 03:23
|
|
|
Dr. Arbitrary posted:Oh, I might pick that up for myself but for my coworkers I'm looking for some structured guides to installing esxi to provisioning a vm and so on. OP has this: http://www.amazon.com/gp/product/0470890800/ref=ox_sc_act_title_2?ie=UTF8&m=ATVPDKIKX0DER That might be more what you are looking for.
|
|
#
?
Apr 29, 2014 04:20
|
|
|
DevNull posted:OP has this: http://www.amazon.com/gp/product/0470890800/ref=ox_sc_act_title_2?ie=UTF8&m=ATVPDKIKX0DER I own that, it's awesome. I'll make what I'm looking for and then maybe it'll be useful to others. I know I've taken a lot from this community. It'll be nice to give back.
|
|
#
?
Apr 29, 2014 05:33
|
|
|
Dilbert, why the hell doesn't View "support" Server 2012 for security/connection servers? This upsets me.
|
|
#
?
May 1, 2014 14:49
|
|
|
Moey posted:Dilbert, why the hell doesn't View "support" Server 2012 for security/connection servers? Coming as soon* I really only noticed that the domain functionality could not be 2012 or higher.
|
|
#
?
May 1, 2014 15:26
|
|
|
Dilbert As gently caress posted:Coming as soon* Yea I noticed that too when I was looking at OS compatibility. I started here over a year ago. So far we have gone from 2000 to 2003..... I just installed the crap on a new security server (2012) in compatibility mode, seems to be working fine. Outside of this being an "unsupported" setup, am I really worried? If I find some strange issues, it doesn't take long to spin up a replacement in 2008 R2.
|
|
#
?
May 1, 2014 15:38
|
|
|
Moey posted:Yea I noticed that too when I was looking at OS compatibility. I started here over a year ago. So far we have gone from 2000 to 2003..... Depends if your using weaker crypo methods on view to combat latency or device capability, it could be an issue. But I doubt it.
|
|
#
?
May 1, 2014 22:33
|
|
|
Need some advice: I've got 3 hosts each with single quad port 1gbit NIC and 1 dual port 10Gbit NIC. In the same rack there are 3 1Gbit 48 port switches (per switch has 2Gbit uplink to our core switch), and in another nearby rack I have two Cisco Cat 4500-X switches (each with 20Gbit uplink to our Core switch). At the switching layer we don't have anything too advanced (I can't even create cross-switch Port Channels (eg virtual port channel)). My qeustion is this: When you only have 2 10Gbit ports to use, how do you load balance/make those links redundant when you can't use port channel?
|
|
#
?
May 2, 2014 01:09
|
|
|
Wicaeed posted:Need some advice: Quick question what licensing level are you on?
|
|
#
?
May 2, 2014 01:19
|
|
|
Dilbert As gently caress posted:Quick question what licensing level are you on? Essentials Plus licensing
|
|
#
?
May 2, 2014 01:19
|
|
|
Oh my god. I found it. The worst VMware environment ever. Full blown company exchange server running off single 1GB non-redundant link to a 4 bay consumer NAS.. 1500mtu.
|
|
#
?
May 2, 2014 01:23
|
|
|
Wicaeed posted:Essentials Plus licensing Use a DVS with NIOC, have your failback to your 2 ports of your QP adapter and 1 backup management and a vMotion off your QP. vMotion works well with NIOC but I find many people like to have vMotion completely segregated onto another nic. You can do off your QP vmnic0 Management-active/vMotion standby vmnic1 vMotion-active/Management-standby Martytoof posted:Oh my god. meh hardly the worse. I've seen desktop PC's with drives dangling 7200RPM 2k drives running in a sudo mirror to host an environment, running freenas without any ZFS. Yeah that's fun Dilbert As FUCK fucked around with this message at 01:31 on May 2, 2014 |
|
#
?
May 2, 2014 01:28
|
|
|
Dilbert As gently caress posted:Coming as soon* Oh shoot, I thought it was just the server it was on. Is there an easy way to fix this or do I need to recreate the domain from scratch. This isn't a problem, I need the practice, but I was hoping to get it working today. I have a question about my plan. My host is on the work network. That way I can access it from my work desktop. All the VMs are on a second network that leads to a second network port that will connect to my test view clients. That keeps me from making a mess out of real stuff. V-center is my weird one. It needs to be on the work network to see the host, right? But at some point the View server will want to see the vcenter server if I understand correctly. Does it automagically do it through VMware tools or will it want an IP I'm already working on adding a second network card onto the vcenter app. Right now I'm trying to do it with a thing called YAST. Does this sound about right for now? Edit: Hey... I don't get errors for eth1 in the vcenter network screen anymore. I think I got it! I wish I'd checked before hitting post.
|
|
#
?
May 2, 2014 01:32
|
|
|
Dilbert As gently caress posted:Use a DVS with NIOC, have your failback to your 2 ports of your QP adapter and 1 backup management and a vMotion off your QP. I thought DVS (Distributed Virtual Switch) was an Enterprise only feature
|
|
#
?
May 2, 2014 01:38
|
|
|
|
| # ? Apr 28, 2024 22:05 |
|
|
Wicaeed posted:I thought DVS (Distributed Virtual Switch) was an Enterprise only feature Somehow I read your post as Enterprise instead of essentials  What you can do is put two vSwithes on those 10G interfaces and limit the outbound traffic. So you can set ISCSI/NFS to XXXXXXX@Kbps/s, vMotion @XXXXXXX@Kbps/s, management XXX@Kbps/s, and use your QP for additional failback. You may not have access to NIOC, but ratelimiting the egress traffic on all hosts is a good way to do a cheap NIOC. I'd probably segregate your vmotion though, that thing is a monster for bandwidth. Dr. Arbitrary posted:Oh shoot, I thought it was just the server it was on. Is there an easy way to fix this or do I need to recreate the domain from scratch. Do you want the lab I built for ICM231, it is basically auto lab for view. I never made a post about it or anything because I am unsure how many people actually would be interested in poo poo like this. Dilbert As FUCK fucked around with this message at 01:51 on May 2, 2014 |
|
#
?
May 2, 2014 01:45
|
|