|
Need advice as we're considering moving to a VMware setup. We currently have a pair of servers ( 1x quad core X3210 2.13G processor, 8GB ram, 4x 7200 rpm sata disks ) running centos and xen. Each server hosts 3 vms, one of which is a mysql instance. This has now become disk bound so we are looking to upgrade to a flash setup. What we are thinking of getting 2x HP DL360p 2 x hex core E5-2620 32 GB ram 365 GB fusion io iodrive2 3 x 600GB SAS disks VMWare essentials (don't need vmotion or the more advanced features ) We would want to be able to run about 8 VMs per server max (3-4 web server , 1 app server, 2 mysql). Does this seem reasonable ?
|
# ¿ Apr 23, 2013 14:55 |
|
|
# ¿ Apr 25, 2024 23:53 |
|
Bob Morales posted:What's the 3x600GB drives for? "slow" storage. Was going to have the os images and database on the flash card and use the drives storing stuff like photos ( of which have gigs and gigs off).
|
# ¿ Apr 23, 2013 15:21 |
|
Corvettefisher posted:So, what happens when a server has a hardware error? What happens if the FusionIO card dies? Is downtime like that acceptable? We already have load balancing across the pair of servers so that's what we use for HA. ( We have a full set of vms running on each physical host). I want to spec it such that one server would to be able to handle the entire load in the event of a failure. The overkill is because we are expanding rapidly and the demand on the services, particularly the DB is likely to increase quite a lot so I want to make sure I don't run out of IO in a years time after spending the money. I considered something like having a flash accelerated NAS but the hosting fees of having an additional two servers would negate the benefits. In addition the existing servers are 5 years old, so I would want to replace them anyway which would mean getting 4 new servers instead of 2.
|
# ¿ Apr 23, 2013 17:23 |
|
Corvettefisher posted:So I take it you are using some application base load balancing and availability? If you are okay with that then yes I am sure your configuration will be much faster than what you have. We have a pair of hardware load balancers upstream of the webserver vms. The mysql instances use floating ips and MMM / heartbeat HA stuff. Corvettefisher posted:IMO, Yes, we don't have our own DC so we pay a per server fee which is significant over the lifetime of the servers. We would need a pair of servers for NAS to avoid a single point of failure so wouldn't we be buying hardware twice anyway ? I take your point that having a storage backend and multiple host machines has a better roadmap for scaling out. Do you have any experience with fusion io or similar PCI express flash cards. My main concern was how well the database stuff would run virtualised even with the flash card. (thanks for all the feedback its appreciated)
|
# ¿ Apr 23, 2013 21:27 |
|
On the subject of vmware licensing, we're going to buy esssentials. Is there any advantage of going with a reseller for this small a purchase or should I just get it from the vmware store?
|
# ¿ May 9, 2013 18:03 |
|
I'm configuring stuff on our new esxi machines now and I'm trying to ensure that a particular partition in one of the virtual machines is aligned with the blocks on the underlying storage. The vmdk will be on local storage formatted with VMFS 5.58 block size 1MB. The storage is a fusion io flash card which is formatted with 512B blocks. I was wanting to create an xfs partition which will be the entire size of the vmdk file. Do I need to bother setting an offset for block alignement or should it just work?
|
# ¿ Jun 5, 2013 16:47 |
|
Its Centos 6.x and I was going to use xfs. I read that if you use mkfs.xfs directly on the disk instead of creating a partition it should be aligned. e.g. mkfs.xfs /dev/sdb edit: added link to source of that info http://www.fhgfs.com/wiki/wikka.php?wakka=PartitionAlignment
|
# ¿ Jun 5, 2013 18:13 |
|
I'm not using raid, the fusion io storage is a PCI card so there's no disk controllers involved at all.
|
# ¿ Jun 5, 2013 18:31 |
|
evol262 posted:Exactly why you should completely ignore the link you posted and just use "parted -a optimal". Following your suggestion I've tried the following code:
http://lists.opensuse.org/opensuse-bugs/2010-05/msg03671.html code:
|
# ¿ Jun 7, 2013 16:28 |
|
evol262 posted:
code:
|
# ¿ Jun 10, 2013 14:32 |
|
evol262 posted:2048 is optimal alignment on top of VMFS-5, yeah. Excellent, thanks for your help.
|
# ¿ Jun 10, 2013 15:39 |
|
demonachizer posted:EDIT: I might have mentioned it before but, primary use for the environment? A file server. I poo poo you not. I was just told the other day that I should try to limit the number of guests that go on it to like 4 or 5... Your company has a lot of a cash and poor planning then ?
|
# ¿ Jun 27, 2013 13:54 |
|
Mr Shiny Pants posted:
The free version of esxi has no limits on physical cpus, The essentials versions have a two socket limit. http://www.vmware.com/products/datacenter-virtualization/vsphere-hypervisor/requirements.html
|
# ¿ Jul 1, 2013 14:01 |
|
zapateria posted:I'm setting up a lab server with the following hardware: If you're running the free version you won't get vmotion, so why don't you just use the ssd and sas as local storage ?
|
# ¿ Jul 1, 2013 14:07 |
|
While folk are talking about certs, I'm looking to get the VCP5-DCV cert. Does anyone know of a qualifying training course in the UK ( or online ) that's not super expensive?Dr. Arbitrary posted:I'm really glad that I've been reading the Scott Lowe book Which Scott Lowe book did you read, this one ? http://www.amazon.co.uk/Mastering-VMware-VSphere-Scott-Lowe/dp/0470890800
|
# ¿ Aug 30, 2013 10:45 |
|
evol262 posted:None of which really matter for "needing Windows guests", to be honest. VMware's ahead because binary translation was incredible and got them ahead in the market really easy. So they're the leaders because nobody else did it and everyone built an ecosystem around VMware. It's a very Microsoft situation, honestly, and VMware's at a point now where they have to fight for their life against competitors, but they're trying to feel their way through (as the vSphere5 licensing debacle 2 years ago illustrated really clearly). They're getting better, but the gap between "what VMware offers" and "what I can get completely for free" is rapidly diminishing. Xen Cloud Platform and oVirt offer 99% of the functionality at 0% of the cost. It's a tough pill for new shops. Your post on open stack was really interesting but you're coming across a bit evangelical. We've just moved away from open source Xen to vmware after 5 years of production use. While in theory they both have the same functionality, the difference is in the quality of the implementation. Our experience was that xen was poorly documented, hard to manage and brittle. Although I believe they fixed it in the most recent version, requiring you to write your own network scripts to get vlan tagging to work was ridiculous. Xen is only cheaper if you have people who can spend all their time learning xen. I'd be interested to hear what you think the market for open stack actually is? If you need arbitary processing power why not just use ec2 or rack space's cloud offering. Why put another layer of abstraction beween you and what your trying to acheive? quote:We're running the same software stack with the same API, and your images can be seamlessly migrated from EC2 to Eucalyptus to RDO to Rackspace with no changes whatsoever. It's the entire point. But rackspace's vm offerings have different specs to amazon which are different to Eucalyptus, so surely you would need to make changes? It also would require you to manage everything yourself and not use any of amazon's managed services like RDS or Dynamo DB. It seems like a solution to a problem that most people who use cloud services don't actually have. Do you have an example of a company utilising open stack in this way?
|
# ¿ Aug 31, 2013 16:07 |
|
evol262 posted:The whole idea here is that there is not "another layer of abstraction" for the end-users. It's exactly the same layer of abstraction that EC2 offers. And to expand on what Mysogynist said, EC2/Rackspace Cloud is really loving expensive. If you're looking at modularizing your environment and you already have significant internal infrastructure, OpenStack can be a win. If you're going to end up deploying on EC2 or Rackspace for production but don't want to pay to run development VMs, and you want an internal environment which matches external, Openstack is aimed directly at you. Thanks, that makes more sense to me as a use case for it. evol262 posted:VMware doesn't always win these comparisons, and the ones they do are because of value-adds, not the excellence or performance of vmkernel itself. VMware is a great one-size-fits-all recommendation for virtualization, but it's not the best at everything. I feel I have a different perspective on this as an end user. I don't really care about the nitty gritty details of which hypervisor is 3% faster at 4k random io under a certain benchmark. For me its how can I use this tool to accomplish stuff the business needs. I am a sysadmin, I have a good understanding networking and how to write network scripts, the time I spent having to gently caress about getting xen to work because they didn't document how to create a virt bridge on a vlan tagged interface is a loss of productivity. The same task in vsphere is typing the vlan number into a box. I didn't even have to rtfm to work out how to do it. My problem was not that the task was complex, It was unnecessarily complex because of a lack of thought on the developers part. Vmware might not be the best at everything, but surely its best at the bit that matters which is getting out the way and letting you actually accomplish work. It also causes a problem in that you need multiple people with an in depth knowledge of this stuff otherwise you get the scenario that I'm getting calls out of hours because none of the rest of the team can trouble shoot it. (always broke when I was on holiday , I think they did it deliberately ) Thats fine in a huge company , but in small to medium business its a real problem, and the expense of hiring extra staff with specialised knowledge totally dwarfs the licensing costs. What do you do when the guys who know how your infastructure works leave, would you not agree its harder to get people with production KVM or Xen experience than vmware? quote:Xen is poorly documented, hard to manage, and can be brittle. So is OpenStack. So is KVM. And Hyper-V. And a few obscure things in VMware. But if you have the in-house expertise, it doesn't matter, and it can be a cost savings. You lose the cost savings when your expensive admins are spending all their time trying to compensate for there being no documentation and the software being brittle instead of doing productive work. Are you not making the same argument that people make when recommending linux on the desktop ? The reason I asked in the first place is that while I recognise there is a market for this It seems to be pretty niche, and limited to people doing massive compute projects who have both the hardware and expertise in house for this to be useful. I don't see how vmware are really losing out by not being involved in it given it's unlikely to be a major revenue generator for anyone selling software.
|
# ¿ Sep 1, 2013 02:49 |
|
adorai posted:These are the same kind of disparaging comments that historically have hurt Linux adoption rates. You are effectively calling him a newb and telling him that poor documentation isn't a bad thing since his team should just know how to do it anyway. In fact, you've implied that he is also a lovely VMware admin for lacking these skills. ^^this The point of the vlan example was that xen 3 required you to do it in a non standard way that wasn't how you would normally do it in RHEL, and they didn't document this anywhere. After looking in the docs for the current version I see that they have since binned this method and allowed you to just create interfaces the normal way. When I was referring to network scripts, I wasn't talking about standard rhel /etc/sysconfig/network-scripts stuff, I agree if you couldn't do that then you are a lovely admin. quote:From Xen 4.1 onwards the xend toolstacks network-bridge script will only reconfigure the host network stack if the network stack does not appear to have been configured already (e.g. no bridges currently exist). This change allows administrators who wish to configure the network stack themselves to do so by default while preserving the existing behaviour for those who do not. Other network-* scripts will still unconditionally reconfigure networking when called by xend. To force xend to never try and reconfigure networking edit /etc/xen/xend-config.sxp and remove any (network-script ...) options. Also having looked at the docs for 4.1, they are now pretty good, so that negates much of my criticism of it. I was only trying to make the point that while ideally you would have lots of highly experienced admins for managing this stuff, in a lot of shops that's not the case and never will be. Anyway, before I started this horrible derail, I was trying to understand why people felt not having a fully openstack compatible vmware offering would cause them to lose market share when the market for openstack and vsphere appear to be quite different. Misogynist posted:VMware needs OpenStack to survive in the age of Amazon Web Services and Google Compute Engine
|
# ¿ Sep 2, 2013 14:41 |
|
evol262 posted:Huh? The pethX/xenbrX stuff was different, but not complicated or arcane, and documentation was fine, with readily available example scripts if you couldn't or didn't want to do it yourself. Xen before 4.1 required a lot of stuff to be done oddly, yeah, but it wasn't so far out of bounds that your typical Linux admin just threw up his hands. Again, if you think that's bad, go look at OpenVSwitch and the ip-netns stuff Yes the docs exist now, they didn't when xen 3 and rhel5 came out, and I've conceded that they've now fixed a lot of the problems. quote:They are not competitors. It's just the "VMware vs the world" argument all over again. Ok thanks, thats a lot clearer.
|
# ¿ Sep 2, 2013 19:42 |
|
Misogynist posted:Have you looked at Logstash? It's not quite Splunk, but it's very free. Thanks for this suggestion, we've implemented it and its fantastic. Pantology posted:vButt Director. This makes my av text a bit sinister
|
# ¿ Sep 6, 2013 16:51 |
|
Latest edition of the Scott Lowe vSphere book is out, mine arrived today reassuringly hefty.
|
# ¿ Nov 15, 2013 13:03 |
|
Dilbert As gently caress posted:but vButt ain't going to happen. Might want to turn off cloud 2 butt before posting/
|
# ¿ Nov 28, 2013 21:19 |
|
I'm thing of reducing the ram our VCenter appliance has assigned from 8G to 4G as we only have two hosts and about 15 vms monitored by it. Is it necessary to manually adjust the java heap values or should it run ok with just reducing the ram ? edit: its 5.1 appliance if that makes any difference jre fucked around with this message at 18:49 on Jan 9, 2014 |
# ¿ Jan 9, 2014 18:46 |
|
Dilbert As gently caress posted:If the linux appliance I'd vouch against going to 4Gb, I've set it to 6Gb on occasions. Also is this update 1 or GA release, update one did help reduce the memory requirements I'd say no less that 6Gb. The appliance is running the VMware-vCenter-Server-Appliance-5.1.0.10100-1123965_OVF10 image. We mainly use it for stats. Can I ask how you determined a 6G minimum ? The knowledge base and ~random blogs~ seem to suggest it would run ok with 4G with a really small deployments like ours. http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=2005086
|
# ¿ Jan 14, 2014 19:33 |
|
Dilbert As gently caress posted:The when using the Webclient to manage the environment I found with less 6GB I got funky errors and random loads to nowhere. How did you reduce the memory size? did you just shrink it in vcenter or did you do the heap size reductions as well ?
|
# ¿ Jan 15, 2014 01:26 |
|
I've an annoying problem with my lab setup. I got a eval copy of workstation 10 from doing the Stanley course and I'm trying to setup a pair of esxi5.5 vms in it. Even though I've added three network interfaces to my vm , the esxi guest will only detect one. I'm not sure if its this bug or something dumb that I'm doing. Anyone had a similar problem ? https://communities.vmware.com/message/2335268
|
# ¿ May 8, 2014 21:57 |
|
jre posted:I've an annoying problem with my lab setup. So turns out this is a weird bug in workstation ... If you install esxi with more than one adaptor added to the VM subsequent nics are detected as the wrong hardware type and don't work 0000:02:01.0 Network controller: Intel Corporation 82545EM Gigabit Ethernet Controller (Copper) [vmnic0] 0000:02:02.0 Network controller: AMD Inc PCnet - Fast 79C971 0000:02:03.0 Network controller: AMD Inc PCnet - Fast 79C971 To get multiple adaptors you must install esxi with only one nic added to vm. When that's fully working shut down esxi, add one adaptor to the vm, boot it up again. Rinse repeat until you have all your adaptors added.
|
# ¿ May 9, 2014 22:24 |
|
Martytoof posted:If they're not on shared storage? I'm looking at my lab vCenter and I don't see the option. I can move from datastore to datastore on the SAME machine, but I dunno. I might be missing something though. Is the vm running ? The option might not show until you shut it down.
|
# ¿ Jul 12, 2014 18:07 |
|
I'm looking to import a couple of virtual box Vms and a xen flat file disk image based vm into esxi. What's the best tool for converting the disk images to a VMware compatible format ?
|
# ¿ Jul 14, 2014 19:16 |
|
esxi 5.5 question, I'm trying to improve memory consolidation on one of our dev servers and I've set Mem.AllocGuestLargePage=0 , but haven't seen any change in host memory usage. Is it enough to restart all the guest OSes or do I need to shut them down all at once? Or is a host reboot needed? It's a standalone host so vmotioning them off then back isn't an option.
|
# ¿ Aug 1, 2014 14:18 |
|
DevNull posted:You just need to power cycle the VM. It doesn't need to be all at once, but you need to actually power off the VM fully. A reboot of the guest won't work. Thanks, I'd read some article on it that talked about "rebooting" but wasn't sure if that referred to guest power cycle or guest reboot.
|
# ¿ Aug 1, 2014 17:42 |
|
Dilbert As gently caress posted:You can also just vmotion stuff around or wait 30 or so minutes for tps to kick in I had applied the setting to the host then left it for a day and there was no change in the host memory usage. Does it definitely work without doing anything to VMs ? Didn't do anything in my case until I power cycled all the VMs.
|
# ¿ Aug 4, 2014 14:23 |
|
edit : ^^^^ Dilbert As gently caress posted:Swap, ballooning, compression that is not being reclaimed. Yeah, I'd read that article before I posted, and one of the first things he does is vmotion everything off and put the host in maintenance mode. quote:So I evacuated a vSphere host using maintenance mode, configured Mem.AllocGuestLargePage to a value of 0, then placed all the VMs back onto the host. Shown below are the before and after results. I'm not saying its not possible, but It did absolutely nothing on my test machine till I power cycled the VMs.
|
# ¿ Aug 5, 2014 15:51 |
|
RevKrule posted:I'm working migrating a lot of computing here to Eucalyptus. It's fantastic for anyone who uses AWS and really it takes maybe an hour to two from blank box to cloud controller (whereas I worked for 9 months getting OpenStack to a state other than "not working" without luck). Have you tried doing something like this ? http://cssoss.wordpress.com/2010/05/04/ueceucalyptus-debugging-instances/
|
# ¿ Sep 16, 2014 16:26 |
|
Got an issue with the impi sensors on one of my hosts. Its an HP DL360p gen8 with the HP esxi 5.1 image on it. If I look in the log its has 4 errors Power supplies are both fine, its just can no longer read any of the statuses from the sensors. Anyone have any ideas how to fix this ?
|
# ¿ Sep 19, 2014 13:07 |
|
luminalflux posted:Get a new PSU from HP? quote:Power supplies are both fine
|
# ¿ Sep 19, 2014 14:26 |
|
Cidrick posted:What's in the IML when you log into iLO? Nothing other than the messages from the install where they tested the supplies.
|
# ¿ Sep 19, 2014 15:39 |
|
adorai posted:if you can separate out those roles from your DC onto separate VMs. As for sizing, I would buy 2x the ram you think you'll need, and I would do an 8 disk raid10 instead of a raid 1 + 6 disk raid 10. In my mind, you would have two disks worth of extra iops sitting idle in your proposed config. Seconding the buy twice as much ram as you think you'll need sentiment. Buy slightly slower processors if necessary to get it in budget.
|
# ¿ Oct 16, 2014 11:12 |
|
Looking at upgrading the ram in some of our hosts and trying to work out exactly what type of DIMMs we have installed without actually opening them first. In vSphere it reports "Memory Type:24". Anyone know what this corresponds to ?
|
# ¿ Dec 1, 2014 18:39 |
|
|
# ¿ Apr 25, 2024 23:53 |
|
cheese-cube posted:Doesn't the host have an iLO/DRAC/IMM web interface that you can login to? It should tell you exactly what DIMMs are installed there. Why yes it does.
|
# ¿ Dec 2, 2014 18:57 |