RE: New Initial Setup Questions (THANK YOU in advance)
On 31.03.2014 22:45, Talk Jesus wrote: Thank you. One last question (most likely)... What difference would it between having a separate storage server as SANS or NAS (especially utilizing something like FreeNAS) for live vm storage + failover? At least in the instance of using CS for a cloud platform. Not sure I follow 100%. As long as that thing can export NFS, it can be anything, FreeNAS or just plain FreeBSD or CentOS etc. -- Sent from the Delta quadrant using Borg technology! Nux! www.nux.ro
RE: New Initial Setup Questions (THANK YOU in advance)
Thank you. One last question (most likely)... What difference would it between having a separate storage server as SANS or NAS (especially utilizing something like FreeNAS) for live vm storage + failover? At least in the instance of using CS for a cloud platform. Thank you. -Original Message- From: Nux! [mailto:n...@li.nux.ro] Sent: Monday, March 31, 2014 5:43 PM To: users@cloudstack.apache.org Subject: RE: New Initial Setup Questions (THANK YOU in advance) On 31.03.2014 20:47, Talk Jesus wrote: >>> The way HA works is ACS will monitor your VMs and HV, if one HV goes > down it will attempt to restart the VMs running on that HV on another > HV, but for this to happen the "new" HV needs to be able to access the > storage those VMs were using and this can only happen if you use > shared storage. Do I make sense? > > Hmm. Not 100% sure I understand, but getting closer :) > > So, "shared storage" would be a separate "storage server", plus at > least 1 extra for a failover? And, if I understand correctly, this > "HA" option is all configured and enabled within the CloudStack > management console itself? > > Sorry if I'm not picking up quickly, but I'm getting it a bit more > clear understanding with a few questions for clarification sake. > > Basically, say I'd like to start a small initial setup to offer cloud > VM's (or popular term "vps / virtual private server" in the cloud). > I'll have one server strictly for the CS management console, I'll have > one HV + 1 extra for failover for CPU/Memory, and then I'll have a > storage server + another for failover. Would this make sense? Yes, this makes sense. The failover SAN/storage is optional. Good luck! -- Sent from the Delta quadrant using Borg technology! Nux! www.nux.ro
RE: New Initial Setup Questions (THANK YOU in advance)
On 31.03.2014 20:47, Talk Jesus wrote: The way HA works is ACS will monitor your VMs and HV, if one HV goes down it will attempt to restart the VMs running on that HV on another HV, but for this to happen the "new" HV needs to be able to access the storage those VMs were using and this can only happen if you use shared storage. Do I make sense? Hmm. Not 100% sure I understand, but getting closer :) So, "shared storage" would be a separate "storage server", plus at least 1 extra for a failover? And, if I understand correctly, this "HA" option is all configured and enabled within the CloudStack management console itself? Sorry if I'm not picking up quickly, but I'm getting it a bit more clear understanding with a few questions for clarification sake. Basically, say I'd like to start a small initial setup to offer cloud VM's (or popular term "vps / virtual private server" in the cloud). I'll have one server strictly for the CS management console, I'll have one HV + 1 extra for failover for CPU/Memory, and then I'll have a storage server + another for failover. Would this make sense? Yes, this makes sense. The failover SAN/storage is optional. Good luck! -- Sent from the Delta quadrant using Borg technology! Nux! www.nux.ro
RE: New Initial Setup Questions (THANK YOU in advance)
>> The way HA works is ACS will monitor your VMs and HV, if one HV goes down it will attempt to restart the VMs running on that HV on another HV, but for this to happen the "new" HV needs to be able to access the storage those VMs were using and this can only happen if you use shared storage. Do I make sense? Hmm. Not 100% sure I understand, but getting closer :) So, "shared storage" would be a separate "storage server", plus at least 1 extra for a failover? And, if I understand correctly, this "HA" option is all configured and enabled within the CloudStack management console itself? Sorry if I'm not picking up quickly, but I'm getting it a bit more clear understanding with a few questions for clarification sake. Basically, say I'd like to start a small initial setup to offer cloud VM's (or popular term "vps / virtual private server" in the cloud). I'll have one server strictly for the CS management console, I'll have one HV + 1 extra for failover for CPU/Memory, and then I'll have a storage server + another for failover. Would this make sense? -Original Message- From: Nux! [mailto:n...@li.nux.ro] Sent: Friday, March 28, 2014 2:24 PM To: users@cloudstack.apache.org Subject: RE: New Initial Setup Questions (THANK YOU in advance) On 28.03.2014 18:09, Talk Jesus wrote: > Thanks Lucian, > >> Ideally you should have separate and redundant (bonded) links for >> management, secondary storage and Internet uplink (public). > > Please explain. I'm thinking, a dedicated Cisco switch for the cloud, > plus another for failover. What do you mean by redundant/bonded links > for management? I think in Cisco-speak that would mean "port channel". Cloudstack has at least 3 types of networks: management, secondary storage, internet uplink. Also take into consideration links to primary network storage if you want it. The network bits are explained in the documentation. In addition to the official docs read the following which I have found helpful: http://shapeblue.com/cloudstack/understanding-cloudstacks-physical-networking-architecture/ http://shankerbalan.net/blog/cloudstack-simple-advanced-network-example/ > >> HA is not possible with local storage > > So local storage = drives for vm "storage" on same HV (cpu/memory > box. I'm curious why HA is not possible on a HV using > CPU/Memory/Storage? I've never heard that before for HA. It looks like we have a language barrier of some sort. :) The way HA works is ACS will monitor your VMs and HV, if one HV goes down it will attempt to restart the VMs running on that HV on another HV, but for this to happen the "new" HV needs to be able to access the storage those VMs were using and this can only happen if you use shared storage. Do I make sense? Lucian -- Sent from the Delta quadrant using Borg technology! Nux! www.nux.ro
RE: New Initial Setup Questions (THANK YOU in advance)
On 28.03.2014 18:09, Talk Jesus wrote: Thanks Lucian, Ideally you should have separate and redundant (bonded) links for management, secondary storage and Internet uplink (public). Please explain. I'm thinking, a dedicated Cisco switch for the cloud, plus another for failover. What do you mean by redundant/bonded links for management? I think in Cisco-speak that would mean "port channel". Cloudstack has at least 3 types of networks: management, secondary storage, internet uplink. Also take into consideration links to primary network storage if you want it. The network bits are explained in the documentation. In addition to the official docs read the following which I have found helpful: http://shapeblue.com/cloudstack/understanding-cloudstacks-physical-networking-architecture/ http://shankerbalan.net/blog/cloudstack-simple-advanced-network-example/ HA is not possible with local storage So local storage = drives for vm "storage" on same HV (cpu/memory box. I'm curious why HA is not possible on a HV using CPU/Memory/Storage? I've never heard that before for HA. It looks like we have a language barrier of some sort. :) The way HA works is ACS will monitor your VMs and HV, if one HV goes down it will attempt to restart the VMs running on that HV on another HV, but for this to happen the "new" HV needs to be able to access the storage those VMs were using and this can only happen if you use shared storage. Do I make sense? Lucian -- Sent from the Delta quadrant using Borg technology! Nux! www.nux.ro
RE: New Initial Setup Questions (THANK YOU in advance)
Thanks Lucian, > Ideally you should have separate and redundant (bonded) links for management, > secondary storage and Internet uplink (public). Please explain. I'm thinking, a dedicated Cisco switch for the cloud, plus another for failover. What do you mean by redundant/bonded links for management? > HA is not possible with local storage So local storage = drives for vm "storage" on same HV (cpu/memory box. I'm curious why HA is not possible on a HV using CPU/Memory/Storage? I've never heard that before for HA. Thanks! -Original Message- From: Nux! [mailto:n...@li.nux.ro] Sent: Friday, March 28, 2014 2:06 PM To: users@cloudstack.apache.org Subject: Re: New Initial Setup Questions (THANK YOU in advance) On 28.03.2014 15:33, Talk Jesus wrote: > One more shot.how's the below for a decent, initial CS setup? > > Server #1 > Purpose: CloudStack Control Panel Install (Management Server) > Qty: 1 > VM Machine (one production, one backup/HA) > Cores: about 2-4 > 4GB RAM It's good; do note you can use both at the same time if you want, just put a load balancer in front of them. > > Server #2 > Purpose: HV (CPU + Memory + Storage) > Qty: 1 plus 1 for failover (HA) > Dual Hexacore CPUs > 128GB RAM > 12 x 2TB SAS (or 6x1TB SSD) RAID10 Played with both SAS and SSD recently, SSD is just so much snappier go for it if you can afford it. > > Server #3 > Purpose: Backup (NAS) > Qty: 1 > Dual Quad Core CPUs > 16GB RAM > 12 x 2TB SATA RAID10 > > > Question: for server #2 (HV), the CPU example is 24 Cores total. So, > technically speaking I can offer a max of 6 VM's with 4 CPU cores > allowanced to each one, correct? Does CS allow "overselling" by any > means on CPU/Memory? You can oversell CPU massively, everyone does, disk and memory are usually the bottlenecks. Overselling memory is trickier, if you want to sell VPS/public cloud I'd advise against it, memory is reasonably cheap. > > Am I missing anything from above initial hardware setup? The network is a very important bit of the setup. Make sure you have quality equipment and redundancy, there's a saying: "your cloud is as good as your network". Ideally you should have separate and redundant (bonded) links for management, secondary storage and Internet uplink (public). > > Anyone have any high availability (HA) setup recommendations for all 3 > servers? I mean does CS have an HA config option within its own > management > console or is HA something done manually outside the CS software, but > within > the o/s on the nodes? Cloudstack can and will perform HA if so instructed, it's "built-in", you don't need to configure anything special in the HV. HA requires: - service offerings with "HA" enabled - shared storage for the above offerings, HA is not possible with local storage Good luck! Lucian -- Sent from the Delta quadrant using Borg technology! Nux! www.nux.ro
Re: New Initial Setup Questions (THANK YOU in advance)
On 28.03.2014 15:33, Talk Jesus wrote: One more shot.how's the below for a decent, initial CS setup? Server #1 Purpose: CloudStack Control Panel Install (Management Server) Qty: 1 VM Machine (one production, one backup/HA) Cores: about 2-4 4GB RAM It's good; do note you can use both at the same time if you want, just put a load balancer in front of them. Server #2 Purpose: HV (CPU + Memory + Storage) Qty: 1 plus 1 for failover (HA) Dual Hexacore CPUs 128GB RAM 12 x 2TB SAS (or 6x1TB SSD) RAID10 Played with both SAS and SSD recently, SSD is just so much snappier go for it if you can afford it. Server #3 Purpose: Backup (NAS) Qty: 1 Dual Quad Core CPUs 16GB RAM 12 x 2TB SATA RAID10 Question: for server #2 (HV), the CPU example is 24 Cores total. So, technically speaking I can offer a max of 6 VM's with 4 CPU cores allowanced to each one, correct? Does CS allow "overselling" by any means on CPU/Memory? You can oversell CPU massively, everyone does, disk and memory are usually the bottlenecks. Overselling memory is trickier, if you want to sell VPS/public cloud I'd advise against it, memory is reasonably cheap. Am I missing anything from above initial hardware setup? The network is a very important bit of the setup. Make sure you have quality equipment and redundancy, there's a saying: "your cloud is as good as your network". Ideally you should have separate and redundant (bonded) links for management, secondary storage and Internet uplink (public). Anyone have any high availability (HA) setup recommendations for all 3 servers? I mean does CS have an HA config option within its own management console or is HA something done manually outside the CS software, but within the o/s on the nodes? Cloudstack can and will perform HA if so instructed, it's "built-in", you don't need to configure anything special in the HV. HA requires: - service offerings with "HA" enabled - shared storage for the above offerings, HA is not possible with local storage Good luck! Lucian -- Sent from the Delta quadrant using Borg technology! Nux! www.nux.ro
New Initial Setup Questions (THANK YOU in advance)
One more shot.how's the below for a decent, initial CS setup? Server #1 Purpose: CloudStack Control Panel Install (Management Server) Qty: 1 VM Machine (one production, one backup/HA) Cores: about 2-4 4GB RAM Server #2 Purpose: HV (CPU + Memory + Storage) Qty: 1 plus 1 for failover (HA) Dual Hexacore CPUs 128GB RAM 12 x 2TB SAS (or 6x1TB SSD) RAID10 Server #3 Purpose: Backup (NAS) Qty: 1 Dual Quad Core CPUs 16GB RAM 12 x 2TB SATA RAID10 Question: for server #2 (HV), the CPU example is 24 Cores total. So, technically speaking I can offer a max of 6 VM's with 4 CPU cores allowanced to each one, correct? Does CS allow "overselling" by any means on CPU/Memory? Am I missing anything from above initial hardware setup? Anyone have any high availability (HA) setup recommendations for all 3 servers? I mean does CS have an HA config option within its own management console or is HA something done manually outside the CS software, but within the o/s on the nodes?
