|
I figure some of you might appreciate a VMware story/rant. I am bringing up some prototype hardware and trying to get ESX running on it. The hardware is really intended to be a laptop, but I really just care about the CPU. However, it means that the storage and network adapters don't work with ESX. My build of ESX is PSODing on boot, but I have no storage or network for logs. My only option is the serial port. My other machines don't have serial, so I have to use a USB to serial adapter on a Windows machine. I boot the ESX machines, and it quickly hit the PSOD. The PSOD triggers a BSOD on my Windows box. I crashed ESX so hard, it killed 2 machines. Let me tell you how much of a joy it is to work with ESX.
|
#
?
Aug 29, 2014 18:41
|
|
|
|
| # ? May 13, 2024 23:16 |
|
|
That's pretty funny. My USB to serial adapter used to cause BSODs all the time. I've never found one that's 100% reliable. I picked an HP Elitebook 8560p two years ago mostly because it had a real serial port and it's the best laptop I've ever used. HP made the new ones look like Macbooks and I believe got rid of the serial ports. I don't think there are any current high end laptops with real serial ports anymore, but I'd be happy to be wrong, because I'm up for another one soon.
|
|
#
?
Aug 29, 2014 18:55
|
|
|
KS posted:That's pretty funny. My USB to serial adapter used to cause BSODs all the time. I've never found one that's 100% reliable. Yeah, I actually tried the USB-serial on my Macbook Air first, and couldn't get any drivers to work. I am not installing linux on an old box that has a serial port. I guess I need to find a laptop with a serial port.
|
|
#
?
Aug 29, 2014 19:03
|
|
|
Cidrick posted:Do any of you guys use Nimble Storage? We're looking to move off of using Hitachi SAN and getting a dedicated storage array solely for VMs, and Nimble seems like a pretty attractive option, but I'm wondering if there's any horror stories out there since it's still a relatively young technology. We have several nimble arrays in production and love them. Initial setup is kind of idiotic, you have to use a machine on the same LAN that uses bonjour with a very specific version or you can't even get an IP address on the damned thing. Once you have that though, it is golden. The VIB for nimble doesn't require a reboot, and installing the vCenter plugin is done easily though the admin console. Very flexible array design too - you can add shelves with more cache, compute, storage without much fuss and no more setup is needed. The CS220(G) series has a limit on expansion shelves, something like 3 total. The others can go much higher. Software is basic but the next release is getting more multiple-tenant capability and catching up feature set wise. Replication being built in with no additional licensing drama is nice too, but we haven't tested that yet.
|
|
#
?
Aug 31, 2014 20:52
|
|
|
KS posted:That's pretty funny. My USB to serial adapter used to cause BSODs all the time. I've never found one that's 100% reliable.
|
|
#
?
Sep 1, 2014 04:07
|
|
|
DevNull posted:Are any of you VMworld people coming down to the south bay at all? I decided to skip VMworld this year, because it is usually pretty boring. VMworld sucked this year. MC Fruit Stripe posted:I think you're losing track of free space there. The existing partition won't be shrinking as the VMDK is expanding, so as Martytoof mentioned, you'll only get 20gb copied into the VMDK before the operation fails. "Oh let me use a program that is judging things of Cylender/disk/header/rotational speed/etc level of understanding and not what the data needs and wants in terms of IOPS or response time!" Do people not understand the layer of abstraction that virtualization provides or just chose to ignore it? Docjowles posted:Interesting announcements from VMware today. Apparently they'll be putting out their own OpenStack distribution next year? Having trouble finding a ton of detail (presumably because of NDA fuckery) but it seems to boil down to official support and easy installation. So you can use all of the OpenStack API's and tooling to build and manage your cloud, and under the hood it's running on vSphere for the hypervisor instead of the more typical KVM or Xen. Which I gather is theoretically possible to do already, but pretty ugly in practice. Been stating this poo poo at VMUG's, classes, and talks I do in the area. VMware won't replace openstack they will just be the go to for management and hypervisors. Dilbert As FUCK fucked around with this message at 06:03 on Sep 1, 2014 |
|
#
?
Sep 1, 2014 05:53
|
|
|
Dilbert As gently caress posted:Why the gently caress do people do this? Do they not understand how VM's play at the SAN, Application, or protocol level? There is literally no one, anywhere, who is incrementally moving files off a partition, defragmenting, shrinking, and repeating.
|
|
#
?
Sep 1, 2014 06:07
|
|
|
MC Fruit Stripe posted:Did you so badly want to pull one back on me that you didn't pay attention to the larger conversation and missed the point? I don't see where I directed that at you, however please point me out where I attacked you specifically. If you read my post I was only commenting on the fact that people actually DO what you spoke of.... Also if you do work in ICM classes or talk at VMUGS you will meet people who do just that. Dilbert As FUCK fucked around with this message at 06:22 on Sep 1, 2014 |
|
#
?
Sep 1, 2014 06:20
|
|
|
Dilbert As gently caress posted:Been stating this poo poo at VMUG's, classes, and talks I do in the area. VMware won't replace openstack they will just be the go to for management and hypervisors. No poo poo. They're part of Openstack. They won't replace anything other than Horizon, and they'll probably leave that, too. But having a sane deployment strategy will put them at the same level as Meraki, but with an extant supply chain. Openstack is hypervisor-agnostic, and there's no such thing as "management" in any VMware-ish sense, except insofar as you can put a provisioning portal on top of Nova (which is what Horizon is). It's a dangerous game to be playing. They're admitting that vCAC isn't up to the job, and investing in an open ecosystem where there's no lock-in at all once you grasp the concepts. I feel like this is a shotgun approach when combined with the Docker announcement -- "we don't know what's happening and what's gonna be big, so we'll just buy into everything" (despite traditional virt not going away). Don't get me wrong. There's a place for Meraki and VMware OpenStack and anyone else who can make the deployment process better. But this is pretty loving far from a slam dunk for VMware and it's not that planned. That said, I'm really curious how they plan on Docker. Build cgroups into vmkernel and add libcontainer to esxi? Run a minimal guest with bindings into vmkernel to balloon memory somehow? I don't know, but it'll be interesting.
|
|
#
?
Sep 1, 2014 06:27
|
|
|
evol262 posted:No poo poo. They're part of Openstack.
|
|
#
?
Sep 1, 2014 06:32
|
|
|
evol262 posted:No poo poo. They're part of Openstack. quote:They won't replace anything other than Horizon, and they'll probably leave that, too. But having a sane deployment strategy will put them at the same level as Meraki, but with an extant supply chain. quote:It's a dangerous game to be playing. They're admitting that vCAC isn't up to the job, and investing in an open ecosystem where there's no lock-in at all once you grasp the concepts. I feel like this is a shotgun approach when combined with the Docker announcement -- "we don't know what's happening and what's gonna be big, so we'll just buy into everything" (despite traditional virt not going away). quote:Don't get me wrong. There's a place for Meraki and VMware OpenStack and anyone else who can make the deployment process better. But this is pretty loving far from a slam dunk for VMware and it's not that planned. I wish I could sit down and have lunch with you so anything I said was not written... Misogynist posted:I'm picturing something more like a VMware-branded take on CoreOS with first-class integration into their other management products. Who even knows, though. Imagine that.... Dilbert As FUCK fucked around with this message at 06:44 on Sep 1, 2014 |
|
#
?
Sep 1, 2014 06:41
|
|
|
Dilbert As gently caress posted:They are the largest corporate contributor, and only said they further support openstack development this con... I'm not gonna say they don't play a part. They do, and it's important, but don't take statements like that without examination. Dilbert As gently caress posted:While Openstack is Hypervisor agnostic, it's a serious pain in the rear end to manage and create environments, unless you have basically a team of linux engineers; it is very obtuse for most enterprises to manage effectively. VMware is going to focus on making the best Hypervisor aware and management for Openstack, you can count on it. I mean, there are a lot of things it's awful at. Which Mirantis does pretty well. And VMware could, but "the best management and hypervisor" is marketing speak. There is no "best" hypervisor, especially since nothing should be FT or even HA (from a VM level) or need to be backed up or do anything VMware is really good at. The "best" hypervisor is the one that works with Nova to start VMs (all of them). KVM, Xen, and vmkernel are better than hyper-v, just because memory page sharing and nested virt are things (and TripleO, aka openstack-on-openstack is a thing). Best management, maybe. This isn't honestly something VMware has excelled at. Managing openstack VMs should be api-driven most of the time, so you're down to managing the install experience, adding the right roles, etc. VMware might be able to nail this, but it's really hard. Mirantis is the best, and they haven't. Canonical has a gui-ish thing that's OK but is new and has low adoption. Rackspace likes chef. We like puppet/foreman. I don't know what VMware's gonna do. There's room, but... Dangerous. I'm not gonna make the "but you can use openstack for free!" argument for a variety of reasons (mostly, enterprises don't care about paying licensing for ease of use and getting people off the street who can use it makes up for investment in specialized candidates), but the cloud model is hard. Once you get used to a certain model and software set, it's hard to add value. Or, in a different sense, the readiness with which companies move between various app servers (jboss v websphere v...) or database servers (not for prod data, obviously, but migrations in dev in rewrites which make it to prod) or SAN vendors should say a lot. If VMware is betting in "management" as "environment install and management", they face a very real possibility of companies using VMware for their initial deploy, then layering more stuff on top of the api they present. That's what they do now. Except Veeam and VAAI don't translate to hyper-v and rhev. Anything built on top of VMware openstack (if anything) will probably he built in the openstack api, and it won't give  if it's Rackspace or VMware or HP or Red Hat.Below... Dilbert As gently caress posted:vCAC owns, it is what vCloud should have been even if it is a "repackage". Even with AWS, which is what I am doing alongside vmware right now, is the meat infront of the dogs mouth; I'd be willing to bet as well as most of the people I have talked with most haven't done cost analysis on it. Most of the companies who rely on IT forgot how to quantify data, AWS mastered the old mainframe stuff and is trying to bring it back. So while we may see a surge into the cloud then a flushback of hybrid/private. VMware will capitalize on Openstack and anything that competes AWS or other cloud venders. VMware needs to pick a product line, basically. Commoditization is still happening. Having their own openstack distro is great if it means they don't pay another vendor for openstack. But ultimately, many people will, because it doesn't matter. ManageIQ and HP's management software and Rightscale really matter. And VMware must have a product in there somewhere. But is it vCAC? Or vCloud Air? Or vRealize? Or plain vCloud? I really don't know. It's impossible to differentiate. And it needs not to be. All their hybrid management should be in one place, and it should be obvious what that place is. I don't at all think traditional virt workloads are going away. We constantly have customers clamor for openstack or openshift, then stare at it with confusion once we deploy it because it just doesn't fit their business model. Some will redo it all. Some won't. Some will move the stuff that makes sense. But once it all settles out, managing the "new" hybrid workload is where it's at. That's the "best management" product I want to see from VMware.
|
|
#
?
Sep 1, 2014 07:44
|
|
|
vCAC and vCloud Director are two fairly different tools that kind of get to a similar place. I'd say vCAC appeals to business minded types and the sorts of folks that order things from web portals. vCloud Director appeals more to dev/test people who might never actually see the UI but instead prefer to drive everything with scripts or Jenkins or something. vCAC is basically what vCloud Director needed at launch for enterprise adopters who care about things like approvals. I won't say it's outright better than vCloud because it does a few things better than vCAC does at the moment and it's also a hell of a lot easier to setup. vCloud Director with vShield (or NSX) provides basically everything a developer might need to get things done without involving anyone else. When we look at vCAC there's a lot of issues with respect to things like adding stuff back into a catalog, managing services other than VMs, and even the integration to AWS. Openstack we're looking at pretty keenly. We've actually recently partnered with Red Hat and have started to deliver Red Hat Openstack for customers. Still a lot of work/not 100% ready for prime time but I think it's time for "early adopters." Also I'm less than thrilled with vCAC's openstack integration.
|
|
#
?
Sep 1, 2014 08:13
|
|
|
evol262 posted:But is it vCAC? Or vCloud Air? Or vRealize? Or plain vCloud? I really don't know. It's impossible to differentiate. And it needs not to be. All their hybrid management should be in one place, and it should be obvious what that place is. Having a simple summary of what each of these does would be really nice. Unfortunately, most of the VMware info around them is a ton of marketing fluff. I bet a majority of VMware employees don't even know the difference.
|
|
#
?
Sep 1, 2014 23:06
|
|
|
DevNull posted:Yeah, I actually tried the USB-serial on my Macbook Air first, and couldn't get any drivers to work. I am not installing linux on an old box that has a serial port. I guess I need to find a laptop with a serial port. I wrangled with this a bit and found a combination that works for me. I've been using a rMBP on mavericks without a crash, but admittedly I don't run faster than 9600 bps. This one uses the FTDI chip AND has pinout lights, woooo http://smile.amazon.com/GearMo%C2%AE-RS-232-Adapter-Indicators-Support/dp/B00AHYJWWG with the FTDI drivers http://www.ftdichip.com/FTDrivers.htm which I learned about from this site http://pbxbook.com/other/mac-tty.html
|
|
#
?
Sep 2, 2014 14:45
|
|
|
Vaporware posted:I wrangled with this a bit and found a combination that works for me. I've been using a rMBP on mavericks without a crash, but admittedly I don't run faster than 9600 bps. That looks like it might be a decent setup. I might give it a try. I ended up finding an old Win XP laptop with a serial port. It works, but would limit what I could do.
|
|
#
?
Sep 2, 2014 15:13
|
|
|
I finally got Veeam... yay! Now the weird part, I'm seeing it's magical for windows VMs, but what about linux based VMs? Any special considerations with Linux and Mysql VMs? Also should I run one backup job to cover the entire infrastructure or should I do individual jobs for each VM? I know the infrastructure level one will add all new VMs and that I will probably need to edit the job to assign proper credentials.
|
|
#
?
Sep 2, 2014 19:38
|
|
|
Do you still need to manually assign VMs to each of your proxy servers, or have they caught up to where PHD Virtual was years ago?
|
|
#
?
Sep 2, 2014 19:43
|
|
|
ghostinmyshell posted:I finally got Veeam... yay! Now the weird part, I'm seeing it's magical for windows VMs, but what about linux based VMs? Any special considerations with Linux and Mysql VMs? For Windows VMs, if you want application-aware processing (VSS snaps from within the guest) you need to select that option. If you need to quiesce a Linux guest, then you need to use a slightly different method that relies on VMware Tools and some pre-freeze and post-thaw scripts. Examples here. As far as what to include - I generally group jobs by role. Windows VMs with application aware processing get one job, AA exclusions get another job (but could go in the AA job if you want), and Linux VMs get their own job. The more systems you have in a job the better dedupe you get (theoretically), but it can also mean more maintenance (exclusions, credentials, etc). YMMV. And no, proxy selection is automatic, and has been since I started using Veeam 1.5-2 years ago (v6.x). Richard Noggin fucked around with this message at 21:18 on Sep 2, 2014 |
|
#
?
Sep 2, 2014 21:16
|
|
|
Thanks, I feel better based on my research so far. I found this page about Linux and database backups so I'll look at this more indepth once I get there. http://www.virtuallifestyle.nl/2013/03/back-up-mysql-on-linux-without-stopping-services-or-dumping-the-database/
|
|
#
?
Sep 2, 2014 21:53
|
|
|
Does Hyper-V schedule virtual processors on a per-core or per-thread basis? I'm wondering how hyperthreading factors into it and how the hypervisor takes away processor time in that case.
|
|
#
?
Sep 3, 2014 14:00
|
|
|
Just curious if anyone has moved to PHD Virtual 8 and wants to share their experience. We use the exporter, which isn't supported on 8 yet, but I'm curious if the product is improved. I've greatly preferred it over Veeam up to this point.
|
|
#
?
Sep 3, 2014 19:15
|
|
|
Hey do ~best practices~ still dictate having a physical DC separate from the rest of the virtual environment? Like are we really worried about that?
|
|
#
?
Sep 3, 2014 19:30
|
|
|
NevergirlsOFFICIAL posted:Hey do ~best practices~ still dictate having a physical DC separate from the rest of the virtual environment? Like are we really worried about that? It doesn't have to be a physical DC, it could be located another virtual environment, but I sleep better at night with a DC outside the main virtual infrastructure for vcenter.
|
|
#
?
Sep 3, 2014 19:39
|
|
|
Side question, How fast does the physical (backup) DC have to be to prevent performance issues? For example I am thinking about setting up AD at home and putting the primary DC on a fast VM and the physical backup would be a spare 15 watt atom powered nettop running WS2008-R2 or WS2012-R2. That would be unusable for GUI use, but fast enough to authenticate against, right? Maybe? I've wanted to setup a DC for AD for a while but absolutely do not trust putting both DCs in the same VM host.
|
|
#
?
Sep 3, 2014 20:02
|
|
|
I've found Server 2008 era documentation from Microsoft that recommends at least one physical DC just in case a "virtualization platform malfunction that affects all host systems that use that platform" wipes out all of your hosts. I'm not sure if they've updated their guidance since then.
|
|
#
?
Sep 3, 2014 20:02
|
|
|
It's just a domain controller, not like you need a lot of horsepower. I just have an old G4p running 2008 R2 in each datacenter. Think you can pick one up for a hundred bucks. It's just active directory and DNS, no real CPU/memory needs there.Hadlock posted:Side question,
|
|
#
?
Sep 3, 2014 21:23
|
|
|
Probably a simple question. Looking to move a number of VMs that are soon out of support on v4.x over to new hardware and on vSphere Enterprise 5.5 What's the 'professional' method/approach/software for such moves?
|
|
#
?
Sep 3, 2014 23:17
|
|
|
Vanilla posted:Probably a simple question.
|
|
#
?
Sep 4, 2014 00:12
|
|
|
Just saw this today. http://blog.mist.io/post/96542374356/one-ui-to-rule-them-all-manage-your-docker-containers
|
|
#
?
Sep 4, 2014 01:13
|
|
|
jaegerx posted:Just saw this today. Openstack already has support. Openshift origin has support (and "proper" openshift will as of 3.0). Panamax does it better. Kubernetes does it better. This is neat and all, but every time I see a post from some startup appealing to the HN crowd and it clearly doesn't understand the  , all I can think is "why?"mist.io posted:And while virtualization targets infrastructure scalability, Docker is ideal for quickly upscaling or downscaling your application, according to your needs. I appreciate what they're trying to do, and it looks neat, but I don't think mist.io will survive (nor will a ton of the other "hitch our startup to a bandwagon" shops). Use kubernetes or panamax or geard
|
|
#
?
Sep 4, 2014 01:51
|
|
|
The tooling built into OpenStack and AWS and Azure are still poo poo for scaling if you need any kind of control. Sending tickets to AWS support to pre-warm an ELB before a marketing push like they're my loving hosting company back in 2002, awesome. Even better when they don't respond to your ticket and the site goes down because the load balancer can't handle the traffic. We ended up just switching to HAProxy because we were tired of dealing with this poo poo. Maybe someday these commodity load balancers will support features that free open-source products supported back in 2000. Vulture Culture fucked around with this message at 03:54 on Sep 4, 2014 |
|
#
?
Sep 4, 2014 03:50
|
|
|
Misogynist posted:The tooling built into OpenStack and AWS and Azure are still poo poo for scaling if you need any kind of control. Sending tickets to AWS support to pre-warm an ELB before a marketing push like they're my loving hosting company back in 2002, awesome. Even better when they don't respond to your ticket and the site goes down because the load balancer can't handle the traffic. I mean, openshift defaults to haproxy, as does openstack's load balancer. And docker can do it fine with fleet or libswarm or geard, but docker requires even more tooling than the "cloud" (some compute service plus a config management service or cloud-init or whatever) does to scale apps, since it has no understanding of multi-node compute scaling, to say nothing of multi-tenant. It's great for app deployment, but does literally nothing for scaling without another tool
|
|
#
?
Sep 4, 2014 04:15
|
|
|
Misogynist posted:Add servers to cluster and vMotion? It doesn't have to be any more complicated than that. Was recommended to use something called Standalone Converter which upgrades the virtual hardware layer and the VM. Offline job though....
|
|
#
?
Sep 5, 2014 19:09
|
|
|
Vanilla posted:Was recommended to use something called Standalone Converter which upgrades the virtual hardware layer and the VM. Offline job though.... Well I am guessing the newer hardware will have incompatible cpus, so you will have to power them off. But the vCenter Converter is more for P2Vs or V2Vs from other hypervisors.
|
|
#
?
Sep 5, 2014 19:14
|
|
|
Moey posted:Well I am guessing the newer hardware will have incompatible cpus, so you will have to power them off. But the vCenter Converter is more for P2Vs or V2Vs from other hypervisors. Target HW will be Intel Ivy Bridge CPUs, bot sure of current but just added the bit above as that was what was suggested.....
|
|
#
?
Sep 5, 2014 20:24
|
|
|
Vanilla posted:Target HW will be Intel Ivy Bridge CPUs, bot sure of current but just added the bit above as that was what was suggested..... Who suggested it? You absolutely do not need to use converter (and an offline convert at that) to do anything. Add to the cluster, vMotion the machines. If you want to upgrade the virtual hardware, power down the VMs, right-click them, and click "upgrade virtual hardware". It doesn't matter at all what the target/source HW will be as long as it's Intel/Intel or AMD/AMD. What matters is the feature compatibility (vmware calls this EVC, I think). But you shouldn't need EVC, because you're going up, not down.
|
|
#
?
Sep 5, 2014 21:12
|
|
|
evol262 posted:Who suggested it? You absolutely do not need to use converter (and an offline convert at that) to do anything. Add to the cluster, vMotion the machines. If you want to upgrade the virtual hardware, power down the VMs, right-click them, and click "upgrade virtual hardware". Fair enough  will revisit
|
|
#
?
Sep 5, 2014 21:27
|
|
|
Do people here present at VMUG's? If so what do you report on? My current session I am looking at is Hypervisors: Do you get what you pay for Storage Deep dive 5.5 And why keeping up to date is important Interested on what others present on at VMUGS and CC's, or how others donate their time to others and feel a need of that "gray area". Last time I did a Q&A about the environment I built and designed. Sadly 6.0 wasn't released so it's back to my 5.x VCDX design. DevNull posted:Having a simple summary of what each of these does would be really nice. Unfortunately, most of the VMware info around them is a ton of marketing fluff. I bet a majority of VMware employees don't even know the difference. That is why Vexperts exist as a "hey do did a bunch for us, have a T shirt!" Dilbert As FUCK fucked around with this message at 03:42 on Sep 6, 2014 |
|
#
?
Sep 6, 2014 03:39
|
|
|
|
| # ? May 13, 2024 23:16 |
|
|
Dilbert As gently caress posted:Sadly 6.0 wasn't released so it's back to my 5.x VCDX design. You shooting for PEX?
|
|
#
?
Sep 6, 2014 03:41
|
|
