|
evol262 posted:I'm also really curious how you're cleaning up the docker images. Does CoreOS come with a cronjob or systemd service for this? It's sort of a non-issue on openstack (except that the root disks tend to be pretty small), but last I checked, docker leaves garbage all over unless you tell "docker rm(i)..." every so often (or really often if you're using docker a lot. There are a few solutions to this problem. There's the cronjob idea (as you suggest) to purge unused images. This doesn't exist AFAIK but it'd be easy to write and if it removed an image that was needed later on then the image would be restored from the Docker private registry when the service was re-launched. Alternatively, if the CoreOS hosts are homogenous cattle and can be recreated on a whim by OpenStack, then it's no bother to periodically kill 'em off and start from scratch with an empty local repository when disc space starts getting tight. The next time a Docker service needs to start on it, it will "docker pull" from the private Docker registry to get the functionality it needs. This works well for 12-factor apps where state is an external service, but doesn't work so well for long-lived VMs. (We call them sheepdogs, because they're somewhere between pets and cattle). The cronjob is the best solution there. However it's rarely been a problem to date because the AUFS-nature of the images means that image footprints are very small so it takes a while to fill up the disc. minato fucked around with this message at 23:35 on Jun 14, 2014 |
# ? Jun 14, 2014 23:33 |
|
|
# ? May 25, 2024 21:33 |
|
This is the first I've heard of Google directly talking about Docker, I guess. They've got some of the best cgroups people in the world, and contributing/driving libcontainer is logical. I'd be totally shocked to see them adopt docker's build system, image registry, etc given their investment in internal cluster tooling, filesystems, etc I also thought docker moved off aufs a few months ago, but I didn't follow the switch that closely. Is it aufs if available, else dm? Or still just aufs? That openstack/Docker mix males sense. Thanks for entertaining my questions. I'm a contributor to nova and also don't get the use case for docker there, but it's basically the same as oVirt (containers run directly on the hypervisor). This is crappy, but works. As a consumer, what could we do to make it (docker support in nova and oVirt) better?
|
# ? Jun 15, 2014 03:08 |
|
You clearly have more insight into this than I do, but I saw this Wired article the other day that sure seems to suggest Google is getting hot and heavy with Docker. ... yes, yes, I just linked a Wired article as a serious source
|
# ? Jun 15, 2014 07:02 |
|
Docker abstracted away the COW filesystem component so that you could choose which one to use. The default uses btrfs but in theory it's possible to swap in any COW filesystem.evol262 posted:As a consumer, what could we do to make it (docker support in nova and oVirt) better? That said, OpenStack already provides a few services above the IaaS layer like LBaaS, DNSaas, etc. These PaaS elements could be Dockerized allowing the user a simple way to add commodity PaaS services right into OpenStack. E.g. a Memcache server or Service Discovery tools. The orchestration, scheduling, service discovery, logging, and monitoring aspects of running a SOA network on Docker are still very immature. Anything that would assist in those aspects of it would be useful. E.g. OpenStack providing an etcd service so that etcd clusters could discover each other.
|
# ? Jun 15, 2014 18:07 |
|
Docjowles posted:You clearly have more insight into this than I do, but I saw this Wired article the other day that sure seems to suggest Google is getting hot and heavy with Docker. Tech media We're all into Docker. Atomic is ostree+Docker, really. Google is moving their environment to docker. Insofar as libcontainer is Docker. But it's ye olde open source problem. We looked at it and said "hey, that's really cool, but..." and dotCloud/Docker looked around and said "hey, we'd have a lot more luck if we stopped hacking around LXC and look at this thing Google did..." and Google said "hey, this model works way better than deploying a whole stack as one container..." And now we're all trying to flesh out what it should be and what it should do, because some of the players (us and canonical, mostly) already have distribution methods and Google doesn't want/need one and... The dotCloud/Docker guys are very much part of the process. And docker isn't going anywhere. The big question is whether other players will use Docker or not. They'll use libcontainer and it'll work with Docker like boxes and virsh and virt-manager and kimchi and nova can all talk to libvirt at the same time. And docker will continue to be a player. But the tech media in general and HN in particular has a really bad tendency to speak in absolutes and overhype everything By the time whatever Atomic (for example) makes it into production in RHEL, it probably won't be directly exposed through Docker, but it'll be on top of libcontainer and Docker will be in @virtualization (but probably not @base). But it might be Docker, too. It's all up in the air.
|
# ? Jun 15, 2014 18:46 |
|
Is there some trick to getting the vSphere Client to format a disk with existing data/partitions as a blank VMFS5 datastore? I threw an SSD into one of my ESXi hosts and I want to use it for host cache bla bla bla, but being the lazy dude I am I didn't bother blanking it ahead of time. vSphere client sees the drive just fine, but it's got Linux partitions and when I try to "Add Storage..." in the Datastores config pane it won't show up as an available device. Do I just need to dd over the MBR or something? I was hoping vClient would give me the option to just blow everything away right there and then. edit: I just ssh'd in and cleared the MBR table and it detects it now. What a pain in the rear end, even if it is an edge case. some kinda jackal fucked around with this message at 02:48 on Jun 29, 2014 |
# ? Jun 29, 2014 02:33 |
|
So, I just got my first project handed to me, it's a fake one (it's just so I get the hang of dealing with projects) but I want to do things right. I'm supposed to design and implement a 2 node SQL server failover cluster on top of a VM. Is there a comparison between the SQL performance with VMWare and Hyper-V? I want to choose the appropriate hypervisor and explain why, and I'd like to know if you guys have any experience in the area or if you know of any study, or performance review.
|
# ? Jun 29, 2014 19:03 |
|
Martytoof posted:Is there some trick to getting the vSphere Client to format a disk with existing data/partitions as a blank VMFS5 datastore? I threw an SSD into one of my ESXi hosts and I want to use it for host cache bla bla bla, but being the lazy dude I am I didn't bother blanking it ahead of time. Is this the thing where you have to format it as vmfs3 first and then 5?
|
# ? Jun 29, 2014 19:12 |
|
Martytoof posted:Is there some trick to getting the vSphere Client to format a disk with existing data/partitions as a blank VMFS5 datastore? I threw an SSD into one of my ESXi hosts and I want to use it for host cache bla bla bla, but being the lazy dude I am I didn't bother blanking it ahead of time. Yeah, especially with SSDs, you have to clear them before you toss 'em in.
|
# ? Jun 29, 2014 20:01 |
|
orange sky posted:So, I just got my first project handed to me, it's a fake one (it's just so I get the hang of dealing with projects) but I want to do things right. I'm supposed to design and implement a 2 node SQL server failover cluster on top of a VM. Is there a comparison between the SQL performance with VMWare and Hyper-V? I want to choose the appropriate hypervisor and explain why, and I'd like to know if you guys have any experience in the area or if you know of any study, or performance review. Honestly these days performance is not the difference between hypervisors. They're all going to be fast or they'd have been driven out of the market. It's more about features, ease of management and which company's licensing works best for you.
|
# ? Jun 29, 2014 20:09 |
|
phosdex posted:Is this the thing where you have to format it as vmfs3 first and then 5? Nah, it's just the thing where ESXi won't allow you to use the disk unless it's "blank"
|
# ? Jun 29, 2014 20:31 |
|
orange sky posted:I want to choose the appropriate hypervisor and explain why "We already have a VMware environment, so I used that" "We already have a Hyper-V environment, so I used that" If neither, pick on cost, future expansion, and how "enterprise" you are (read: likely to get into all the add-ons for either product).
|
# ? Jun 30, 2014 14:42 |
|
orange sky posted:So, I just got my first project handed to me, it's a fake one (it's just so I get the hang of dealing with projects) but I want to do things right. I'm supposed to design and implement a 2 node SQL server failover cluster on top of a VM. Is there a comparison between the SQL performance with VMWare and Hyper-V? I want to choose the appropriate hypervisor and explain why, and I'd like to know if you guys have any experience in the area or if you know of any study, or performance review. In addition to what evol262 said, you're missing the point of the exercise. You're basically starting with should I run SQL on HP servers or Dell servers. It doesn't loving matter. If asked someone to do this exercise and they put that much thought into the hypervisor I would be pissed, as they missed the point. I wanted someone to design and implement a SQL cluster and all it's associated software and design requirements. I also want someone to put together a reasonable project plan and timeline. Scope, Objectives, Resources needed, Schedule etc. gently caress a hypervisor. How's the saying go? You're focusing on a single tree, and not seeing the forest... whatever, you get the point I hope. edit: In retrospect maybe I'm being too hasty and you just came in here to ask about hypervisors and have a handle on the rest of the project. There may or may not be 4 or 6% difference between the two in very specific circumstances, either way it really won't matter and you'll use whatever hypervisor technology is already in place. I know we wouldn't bring up a Hyper-V cluster for a single app and single digit theoretical performance increases. So it's not really a question to deal with to be honest. Requirements should just be 2 appropriately sized virtual machines that can be allocated from the existing environment. skipdogg fucked around with this message at 23:33 on Jun 30, 2014 |
# ? Jun 30, 2014 23:27 |
|
A new vSphere beta just launched yesterday: https://communities.vmware.com/community/vmtn/vsphere-beta It is available to anyone, but not "open" because you still have to sign the NDA and can't talk about it in public.
|
# ? Jul 1, 2014 15:36 |
|
skipdogg posted:In addition to what evol262 said, you're missing the point of the exercise. You're basically starting with should I run SQL on HP servers or Dell servers. It doesn't loving matter. Ok, got it! Yeah I wasn't really extremely concerned about it, I just wanted to know if you guys knew about something I didn't that really hindered the performance when using VMWare or something. I think the VMs they're going to give me are on Hyper-V so I'll just take that.
|
# ? Jul 1, 2014 16:21 |
|
My company is going to let customers reserve CPU and Memory resources on their own.
|
# ? Jul 1, 2014 20:52 |
|
You poor bastard.
|
# ? Jul 1, 2014 20:59 |
|
Does Host Cache not suppland individual swap files stored with the VM? I was hoping that if I dedicated a 120GB SSD to host cache I could bypass the swap files stored on slower storage. I told the host to store swapfiles on the SSD, but it creates the swap file on the SSD in addition to the host cache swap files it creates. I guess I'm not really sure what the host cache is for then. I'm not sure I'm articulating this right so hopefully it makes sense to someone.
|
# ? Jul 1, 2014 21:06 |
|
mAlfunkti0n posted:My company is going to let customers reserve CPU and Memory resources on their own. Why don't they just let them make requests and your company handles it later? That poo poo is a can of worms..
|
# ? Jul 1, 2014 21:07 |
|
orange sky posted:Why don't they just let them make requests and your company handles it later? That poo poo is a can of worms.. I've stopped asking why. I've stopped asking any intelligent questions.
|
# ? Jul 1, 2014 21:08 |
|
DevNull posted:A new vSphere beta just launched yesterday: https://communities.vmware.com/community/vmtn/vsphere-beta So is this a Beta of something like vSphere 6.0 or some lesser intermediate product?
|
# ? Jul 1, 2014 21:50 |
|
FISHMANPET posted:So is this a Beta of something like vSphere 6.0 or some lesser intermediate product? Hmm, that is weird that beta signup doesn't have that. I guess it is safe to publicly say it is a major release.
|
# ? Jul 1, 2014 22:15 |
|
DevNull posted:Hmm, that is weird that beta signup doesn't have that. I guess it is safe to publicly say it is a major release. That's because it was a user sponsered post, that has been removed from blogs.vmware.com. Talking about the Beta is still under NDA unless you are in the forum, Hardware, network, or storage vendor. It was a misstep on vmware's part FISHMANPET posted:So is this a Beta of something like vSphere 6.0 or some lesser intermediate product? I don't know if I can echo the post as it violated NDA. However look up what Beta means in software progression and find out if it means anything. Dilbert As FUCK fucked around with this message at 01:02 on Jul 2, 2014 |
# ? Jul 2, 2014 00:59 |
|
mAlfunkti0n posted:My company is going to let customers reserve CPU and Memory resources on their own. Can you use resource pools to keep things from getting too far out of hand? I haven't played with reservations too much, what's the worst that can happen?
|
# ? Jul 2, 2014 01:38 |
|
Dr. Arbitrary posted:Can you use resource pools to keep things from getting too far out of hand? I haven't played with reservations too much, what's the worst that can happen? vCAC is a much friendly front end solution.
|
# ? Jul 2, 2014 01:42 |
|
This https://blogs.vmware.com/vsphere/2014/06/now-open-vmware-vsphere-beta-program.html was posted by a VMware employee and it's still up. Here's Duncan Epping being very public about it: http://www.yellow-bricks.com/2014/07/01/public-vsphere-beta-sign-provide-feedback-now/ It's not a misstep at all on VMware's part. This is very deliberate and just the first time they've done this. Probably weren't happy with financials or something haha. If you go sign up with an account (can use an existing communities account) you can get all the information you'd ever want to know including what version this will most likely be released as.
|
# ? Jul 2, 2014 01:57 |
|
1000101 posted:This https://blogs.vmware.com/vsphere/2014/06/now-open-vmware-vsphere-beta-program.html Sorry but is it really to that point where VMware is actively leaking beta intel to counter hyper-v? 6.0 is cool as poo poo, no idea how can I can go on it, but wow I love it thus far, aside from the SQL 2012 support.... E: Also if this is so public what can we leak about 6.0? Some of the poo poo is so unbelievably cool. Dilbert As FUCK fucked around with this message at 02:03 on Jul 2, 2014 |
# ? Jul 2, 2014 02:01 |
|
Dilbert As gently caress posted:Sorry but is it really to that point where VMware is actively leaking beta intel to counter hyper-v? It's probably more like "we've got a lot of cool poo poo in this beta and we want to pump up our stock/hype things up going into vmworld." Also may be a genuine experiment on their part since the last few releases have been pretty light on the QA. If you open it up to a large crowd the hope might be to get a lot of meaningful feedback to button it up before VMworld.
|
# ? Jul 2, 2014 02:05 |
|
1000101 posted:It's probably more like "we've got a lot of cool poo poo in this beta and we want to pump up our stock/hype things up going into vmworld." Maybe the vCenter installer won't poo poo the bed when you hit 'simple install' this time.
|
# ? Jul 2, 2014 02:10 |
|
1000101 posted:It's probably more like "we've got a lot of cool poo poo in this beta and we want to pump up our stock/hype things up going into vmworld." That figures, the NDA was tight this round from what I've seen. I think Hyper-V is giving them a bit of a chase. U2 fixed a bunch of poo poo they really listened which is nice! I just hope they don't cut of the VCDX 5 design so short.... I mean I don't mind climbing the ladder but gently caress man... madsushi posted:Maybe the vCenter installer won't poo poo the bed when you hit 'simple install' this time. If you don't have an odbc connection I have "NDA news" for you!
|
# ? Jul 2, 2014 02:16 |
|
Dr. Arbitrary posted:Can you use resource pools to keep things from getting too far out of hand? I haven't played with reservations too much, what's the worst that can happen? I can and sometimes do implement things on my own to curb pain points. In this case though, the VM has to be provisioned and then stuck into a pool, which the system wont do. If we have even a small percentage of the servers all of a sudden (and I have a feeling this is going to happen) have reservation changes, other VMs will starve for resources and when there is a host failure there is a darn good chance stuff just won't failover. The problem with our scenario is that we are not Amazon with insane resources (though we are a large company). There is no real accountability, and enforcement, well we just aren't allowed to do that. I've shown some of the server "owners" that their performance problems stem from the fact that they have eight vCPUs and need two, giving them real data and letting them test as well, only to be told "well we WANT eight". The bigger problem is, everyone acts like that .. so you can imagine the waste. Oh well, I guess its more OT for me when stuff blows up. I' have voiced my concerns, so they can't blame me!
|
# ? Jul 2, 2014 02:29 |
|
1000101 posted:It's probably more like "we've got a lot of cool poo poo in this beta and we want to pump up our stock/hype things up going into vmworld." Mostly the testing part. We do a ton of QA in house, but there have been some horror stories in the past because the beta was so small. Customers are the best at finding weird bugs. The NDA is just to keep people from going out and posting articles and such before it actually releases.
|
# ? Jul 2, 2014 02:58 |
|
DevNull posted:Mostly the testing part. We do a ton of QA in house, but there have been some horror stories in the past because the beta was so small. Customers are the best at finding weird bugs. The NDA is just to keep people from going out and posting articles and such before it actually releases. Your MSQL migrations to ....THAT. could use some attention. I mean I can gently caress my way through it but you all get paid the mega bucks right?
|
# ? Jul 2, 2014 03:01 |
|
Dilbert As gently caress posted:
You'll probably be able to defend on 5 up till about a year after 6 is released. This is mostly because panels are typically twice a year in each major region. Have you started building your design yet?
|
# ? Jul 2, 2014 03:55 |
|
1000101 posted:You'll probably be able to defend on 5 up till about a year after 6 is released. This is mostly because panels are typically twice a year in each major region. quote:Have you started building your design yet? Designed and sold a few people + a VCDX on it. It's in the writing stages now.... I actually did a slight mock defense at the last VMug I spoke at because I was curious. Outlined a 40 page(oh gently caress it needs to be longer don't it) + visio diagram detailed doc on it; but I don't think it's good enough. I resorted back to a 4 host design + dual san design. I want to go more complex but everyone advises against it for some reason... So yeah I put off that storage OP in the mean time. Also there is a substantial motivation to surprise #128 on it. Dilbert As FUCK fucked around with this message at 04:11 on Jul 2, 2014 |
# ? Jul 2, 2014 04:02 |
|
Dilbert As gently caress posted:
Complexity is frequently the enemy of reliability.
|
# ? Jul 2, 2014 06:17 |
|
1000101 posted:Complexity is frequently the enemy of reliability. Yeah that's the point I drove home at the Hampton Roads VMUG, It's easy to build a complex environment, but hard to build a simple environment. http://www.vhersey.com/2014/06/hrvmug-meeting-report/ Why is FINALLY why I got our storage accelerator on a VM and got it backed up via VDP.... Complexity is the Achilles's heal in any design because you always overlook the 'simple stuff'.
|
# ? Jul 2, 2014 06:20 |
|
Ugh I really don't feel like signing up for a beta just to see if one feature is there. Anyone in, please blink once if VUM still requires windows or twice if I can fully switch to using VCSA.
|
# ? Jul 2, 2014 15:05 |
|
Dilbert As gently caress posted:Your MSQL migrations to ....THAT. could use some attention. I mean I can gently caress my way through it but you all get paid the mega bucks right? Post this in the official beta forum. I don't know who exactly in is charge of that, but I know it is the type of ting that our CTO is really looking into making easier. He made a bunch of engineers go though some training recently just to see how bad our upgrade process is. He went though it himself as well. If you point out specifics, they are much more likely to get addressed. Or PM me details and I can see who I can find to look into it. If you want to use the communities forums to message me: https://communities.vmware.com/people/TheDevNull
|
# ? Jul 2, 2014 15:43 |
|
|
# ? May 25, 2024 21:33 |
|
DevNull posted:Post this in the official beta forum. I don't know who exactly in is charge of that, but I know it is the type of ting that our CTO is really looking into making easier. He made a bunch of engineers go though some training recently just to see how bad our upgrade process is. He went though it himself as well. If you point out specifics, they are much more likely to get addressed. Yeah I'll upload more logs, might make a thread on the beta forum
|
# ? Jul 2, 2014 15:48 |