You would setup three servers first in hyperconverged using either replica 3 or replica 3 arbiter 1 then add your fourth host afterward as a compute only host that can host vms but does not participate in glusterfs storage.
On Sun, Jul 19, 2020 at 3:12 PM David White via Users <users@ovirt.org> wrote: > Thank you. > So to make sure I understand what you're saying, it sounds like if I need > 4 nodes (or more), I should NOT do a "hyperconverged" installation, but > should instead prepare Gluster separately from the oVirt Manager > installation. Do I understand this correctly? > > If that is the case, can I still use some of the servers for dual purposes > (Gluster + oVirt Manager)? I'm most likely going to need more servers for > the storage than I will need for the RAM & CPU, which is a little bit > opposite of what you wrote (using 3 servers for Gluster and adding > additional nodes for RAM & CPU). > > > Sent with ProtonMail Secure Email. > > ‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐ > On Sunday, July 19, 2020 9:57 AM, Strahil Nikolov via Users < > users@ovirt.org> wrote: > > > Hi David, > > > > > it's a little bit different. > > > > > Ovirt supports 'replica 3' (3 directories hsot the same content) or > 'replica 3 arbiter 1' (2 directories host same data, third directory > contains metadata to prevent split brain situations) volumes. > > > > > If you have 'replica 3' it is smart to keep the data on separate hosts, > although you can keep it on the same host (but then you should use no > replica and oVirt's Single node setup). > > > > > When you extend , yoou need to add bricks (fancy name for a directory) > in the x3 count. > > > > > If you wish that you want to use 5 nodes, you can go with 'replica 3 > arbiter 1' volume, where ServerA & ServerB host data and ServerC host only > metadata (arbiter). Then you can extend and for example ServerC can host > again metadata while ServerD & ServerE host data for the second replica set. > > > > > You can even use only 3 servers for Gluster , while much more systems as > ovirt nodes (CPU & RAM) to host VMs. > > In case of a 4 node setup - 3 hosts have the gluster data and the 4th - > is not part of ths gluster, just hosting VMs. > > > > > Best Regards, > > Strahil Nikolov > > > > > На 19 юли 2020 г. 15:25:10 GMT+03:00, David White via Users > users@ovirt.org написа: > > > > > > Thanks again for explaining all of this to me. > > > Much appreciated. > > > Regarding the hyperconverged environment, > > > reviewing > https://www.ovirt.org/documentation/gluster-hyperconverged/chap-Deploying_Hyperconverged.html > , > > > it appears to state that you need, exactly, 3 physical servers. > > > Is it possible to run a hyperconverged environment with more than 3 > > > physical servers? > > > Because of the way that the gluster triple-redundancy works, I knew > > > that I would need to size all 3 physical servers' SSD drives to store > > > 100% of the data, but there's a possibility that 1 particular (future) > > > customer is going to need about 10TB of disk space. > > > For that reason, I'm thinking about what it would look like to have 4 > > > or even 5 physical servers in order to increase the total amount of > > > disk space made available to oVirt as a whole. And then from there, I > > > would of course setup a number of virtual disks that I would attach > > > back to that customer's VM. > > > So to recap, if I were to have a 5-node Gluster Hyperconverged > > > environment, I'm hoping that the data would still only be required to > > > replicate across 3 nodes. Does this make sense? Is this how data > > > replication works? Almost like a RAID -- add more drives, and the RAID > > > gets expanded? > > > Sent with ProtonMail Secure Email. > > > ‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐ > > > On Tuesday, June 23, 2020 4:41 PM, Jayme jay...@gmail.com wrote: > > > > > > > > Yes this is the point of hyperconverged. You only need three hosts to > > > > setup a proper hci cluster. I would recommend ssds for gluster > storage. > > > > You could get away with non raid to save money since you can do > replica > > > > three with gluster meaning your data is fully replicated across all > > > > three hosts. > > > > > > > > On Tue, Jun 23, 2020 at 5:17 PM David White via Users > > > > users@ovirt.org wrote: > > > > > > > > > Thanks. > > > > > I've only been considering SSD drives for storage, as that is what > > > > > I currently have in the cloud. > > > > > > > > > > > > > > > > > > > > I think I've seen some things in the documents about oVirt and > > > > > gluster hyperconverged. > > > > > > > > > > Is it possible to run oVirt and Gluster together on the same > > > > > hardware? So 3 physical hosts would run CentOS or something, and I > > > > > would install oVirt Node + Gluster onto the same base host OS? If > so, > > > > > then I could probably make that fit into my budget. > > > > > > > > > > > > > > > > > > > > Sent with ProtonMail Secure Email. > > > > > > > > > ‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐ > > > > > On Monday, June 22, 2020 1:02 PM, Strahil Nikolov via Users > > > > > users@ovirt.org wrote: > > > > > > > > > > > > > > > > > > > > > Hey David, > > > > > > > > > > keep in mind that you need some big NICs. > > > > > > I started my oVirt lab with 1 Gbit NIC and later added 4 > > > > > > dual-port 1 Gbit NICs and I had to create multiple gluster > volumes and > > > > > > multiple storage domains. > > > > > > > > > > > Yet, windows VMs cannot use software raid for boot devices, thus > > > > > > it's a pain in the @$$. > > > > > > > > > > > I think that optimal is to have several 10Gbit NICs (at least 1 > > > > > > for gluster and 1 for oVirt live migration). > > > > > > > > > > > Also, NVMEs can be used as lvm cache for spinning disks. > > > > > > > > > > Best Regards, > > > > > > Strahil Nikolov > > > > > > > > > > На 22 юни 2020 г. 18:50:01 GMT+03:00, David White > > > > > > dmwhite...@protonmail.com написа: > > > > > > > > > > > > > > > > > > > > > > > > For migration between hosts you need a shared storage. SAN, > > > > > > > > Gluster, > > > > > > > > > > > > > CEPH, NFS, iSCSI are among the ones already supported (CEPH > > > > > > > > is a little > > > > > > > > > > > > > bit experimental). > > > > > > > > > > > Sounds like I'll be using NFS or Gluster after all. > > > > > > > Thank you. > > > > > > > > > > > > The engine is just a management layer. KVM/qemu has that > > > > > > > > option a > > > > > > > > > > > > > long time ago, yet it's some manual work to do it. > > > > > > > > Yeah, this environment that I'm building is expected to grow > > > > > > > > over time > > > > > > > > > > > > > (although that growth could go slowly), so I'm trying to > > > > > > > > architect > > > > > > > > > > > > > things properly now to make future growth easier to deal > > > > > > > > with. I'm also > > > > > > > > > > > > > trying to balance availability concerns with budget > > > > > > > > constraints > > > > > > > > > > > > > starting out. > > > > > > > > > > > Given that NFS would also be a single point of failure, I'll > > > > > > > probably > > > > > > > > > > > > go with Gluster, as long as I can fit the storage requirements > > > > > > > into the > > > > > > > > > > > > overall budget. > > > > > > > Sent with ProtonMail Secure Email. > > > > > > > ‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐ > > > > > > > On Monday, June 22, 2020 6:31 AM, Strahil Nikolov via Users > > > > > > > users@ovirt.org wrote: > > > > > > > > > > > > На 22 юни 2020 г. 11:06:16 GMT+03:00, David White via > > > > > > > > usersus...@ovirt.org написа: > > > > > > > > > > > > > Thank you and Strahil for your responses. > > > > > > > > > They were both very helpful. > > > > > > > > > > > > > > I think a hosted engine installation VM wants 16GB RAM > > > > > > > > > > configured > > > > > > > > > > > > > > > though I've built older versions with 8GB RAM. > > > > > > > > > > For modern VMs CentOS8 x86_64 recommends at least 2GB for > > > > > > > > > > a host. > > > > > > > > > > > > > > > CentOS7 was OK with 1, CentOS6 maybe 512K. > > > > > > > > > > The tendency is always increasing with updated OS > > > > > > > > > > versions. > > > > > > > > > > > > > > > > > > > > > > > > > > Ok, so to clarify my question a little bit, I'm trying to > > > > > > > > > figure > > > > > > > > > > > > > > out > > > > > > > > > > > > > how much RAM I would need to reserve for the host OS (or > > > > > > > > > oVirt > > > > > > > > > > > > > > Node). > > > > > > > > > > > > > I do recall that CentOS / RHEL 8 wants a minimum of 2GB, so > > > > > > > > > perhaps > > > > > > > > > > > > > > that would suffice? > > > > > > > > > And then as you noted, I would need to plan to give the > > > > > > > > > engine > > > > > > > > > > > > > > 16GB. > > > > > > > > > > > > I run my engine on 4Gb or RAM, but i have no more than 20 > > > > > > > > VMs, the > > > > > > > > > > > > > larger the setup - the more ram for the engine is needed. > > > > > > > > > > > > > > My minimum ovirt systems were mostly 48GB 16core, but > > > > > > > > > > most are > > > > > > > > > > > > > > > now > > > > > > > > > > > > > > 128GB 24core or more. > > > > > > > > > > > > > But this is the total amount of physical RAM in your > > > > > > > > > systems, > > > > > > > > > > > > > > correct? > > > > > > > > > > > > > Not the amount that you've reserved for your host OS?I've > > > > > > > > > spec'd > > > > > > > > > > > > > > out > > > > > > > > > > > > > some hardware, and am probably looking at purchasing two > > > > > > > > > PowerEdge > > > > > > > > > > > > > > R820's to start, each with 64GB RAM and 32 cores. > > > > > > > > > > > > > > While ovirt can do what you would like it to do > > > > > > > > > > concerning a > > > > > > > > > > > > > > > single > > > > > > > > > > > > > > user interface, but with what you listed, > > > > > > > > > > you're probably better off with just plain KVM/qemu and > > > > > > > > > > using > > > > > > > > > > > > > > > virt-manager for the interface. > > > > > > > > > > > > > Can you migrate VMs from 1 host to another with > > > > > > > > > virt-manager, and > > > > > > > > > > > > > > can > > > > > > > > > > > > > you take snapshots? > > > > > > > > > If those two features aren't supported by virt-manager, > > > > > > > > > then that > > > > > > > > > > > > > > would > > > > > > > > > > > > > almost certainly be a deal breaker. > > > > > > > > > > > > The engine is just a management layer. KVM/qemu has that > > > > > > > > option a > > > > > > > > > > > > > long time ago, yet it's some manual work to do it. > > > > > > > > > > > > > Come to think of it, if I decided to use local storage on > > > > > > > > > each of > > > > > > > > > > > > > > the > > > > > > > > > > > > > physical hosts, would I be able to migrate VMs? > > > > > > > > > Or do I have to use a Gluster or NFS store for that? > > > > > > > > > > > > For migration between hosts you need a shared storage. SAN, > > > > > > > > Gluster, > > > > > > > > > > > > > CEPH, NFS, iSCSI are among the ones already supported (CEPH > > > > > > > > is a little > > > > > > > > > > > > > bit experimental). > > > > > > > > > > > > > ‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐ > > > > > > > > > On Sunday, June 21, 2020 5:58 PM, Edward Berger > > > > > > > > > edwber...@gmail.com > > > > > > > > > > > > > > wrote: > > > > > > > > > > > > > > While ovirt can do what you would like it to do > > > > > > > > > > concerning a > > > > > > > > > > > > > > > single > > > > > > > > > > > > > > user interface, but with what you listed, > > > > > > > > > > you're probably better off with just plain KVM/qemu and > > > > > > > > > > using > > > > > > > > > > > > > > > virt-manager for the interface. > > > > > > > > > > > > > > Those memory/cpu requirements you listed are really tiny > > > > > > > > > > and I > > > > > > > > > > > > > > > wouldn't recommend even trying ovirt on such challenged > > > > > > > > > > systems. > > > > > > > > > > > > > > > I would specify at least 3 hosts for a gluster > > > > > > > > > > hyperconverged > > > > > > > > > > > > > > > system, > > > > > > > > > > > > > > and a spare available that can take over if one of the > > > > > > > > > > hosts > > > > > > > > > > > > > > > dies. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > I think a hosted engine installation VM wants 16GB RAM > > > > > > > > > > configured > > > > > > > > > > > > > > > though I've built older versions with 8GB RAM. > > > > > > > > > > For modern VMs CentOS8 x86_64 recommends at least 2GB for > > > > > > > > > > a host. > > > > > > > > > > > > > > > CentOS7 was OK with 1, CentOS6 maybe 512K. > > > > > > > > > > The tendency is always increasing with updated OS > > > > > > > > > > versions. > > > > > > > > > > > > > > > > > > > > > > > > > > > My minimum ovirt systems were mostly 48GB 16core, but > > > > > > > > > > most are > > > > > > > > > > > > > > > now > > > > > > > > > > > > > > 128GB 24core or more. > > > > > > > > > > > > > > ovirt node ng is a prepackaged installer for an oVirt > > > > > > > > > > hypervisor/gluster host, with its cockpit interface you > > > > > > > > > > can > > > > > > > > > > > > > > > create and > > > > > > > > > > > > > > install the hosted-engine VM for the user and admin web > > > > > > > > > > interface. Its > > > > > > > > > > > > > > very good on enterprise server hardware with lots of > > > > > > > > > > RAM,CPU, and > > > > > > > > > > > > > > > DISKS. > > > > > > > > > > > > > > On Sun, Jun 21, 2020 at 4:34 PM David White via Users > > > > > > > > > > users@ovirt.org wrote: > > > > > > > > > > > > > > > I'm reading through all of the documentation at > > > > > > > > > > > https://ovirt.org/documentation/, and am a bit > > > > > > > > > > > overwhelmed with > > > > > > > > > > > > > > > > all of > > > > > > > > > > > > > > > the different options for installing oVirt. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > My particular use case is that I'm looking for a way to > > > > > > > > > > > manage > > > > > > > > > > > > > > > > VMs > > > > > > > > > > > > > > > on multiple physical servers from 1 interface, and be > > > > > > > > > > > able to > > > > > > > > > > > > > > > > deploy > > > > > > > > > > > > > > > new VMs (or delete VMs) as necessary. Ideally, it would > > > > > > > > > > > be > > > > > > > > > > > > > > > > great if I > > > > > > > > > > > > > > > could move a VM from 1 host to a different host as > > > > > > > > > > > well, > > > > > > > > > > > > > > > > particularly > > > > > > > > > > > > > > > in the event that 1 host becomes degraded (bad HDD, bad > > > > > > > > > > > processor, > > > > > > > > > > > > > > > etc...) > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > I'm trying to figure out what the difference is between > > > > > > > > > > > an > > > > > > > > > > > > > > > > oVirt > > > > > > > > > > > > > > > Node and the oVirt Engine, and how the engine differs > > > > > > > > > > > from the > > > > > > > > > > > > > > > > Manager. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > I get the feeling that `Engine` = `Manager`. Same > > > > > > > > > > > thing. I > > > > > > > > > > > > > > > > further > > > > > > > > > > > > > > > think I understand the Engine to be essentially > > > > > > > > > > > synonymous with > > > > > > > > > > > > > > > > a > > > > > > > > > > > > > > > vCenter VM for ESXi hosts. Is this correct? > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > If so, then what's the difference between the > > > > > > > > > > > `self-hosted` vs > > > > > > > > > > > > > > > > the > > > > > > > > > > > > > > > `stand-alone` engines? > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > oVirt Engine requirements look to be a minimum of 4GB > > > > > > > > > > > RAM and > > > > > > > > > > > > > > > > 2CPUs. > > > > > > > > > > > > > > > oVirt Nodes, on the other hand, require only 2GB RAM. > > > > > > > > > > > Is this a requirement just for the physical host, or is > > > > > > > > > > > that > > > > > > > > > > > > > > > > how > > > > > > > > > > > > > > > much RAM that each oVirt node process requires? In > > > > > > > > > > > other words, > > > > > > > > > > > > > > > > if I > > > > > > > > > > > > > > > have a physical host with 12GB of physical RAM, will I > > > > > > > > > > > only be > > > > > > > > > > > > > > > > able to > > > > > > > > > > > > > > > allocate 10GB of that to guest VMs? How much of that > > > > > > > > > > > should I > > > > > > > > > > > > > > > > dedicated > > > > > > > > > > > > > > > to the oVirt node processes? > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Can you install the oVirt Engine as a VM onto an > > > > > > > > > > > existing oVirt > > > > > > > > > > > > > > > > Node? And then connect that same node to the Engine, > > > > > > > > > > > once the > > > > > > > > > > > > > > > > Engine is > > > > > > > > > > > > > > > installed? > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Reading through the documentation, it also sounds like > > > > > > > > > > > oVirt > > > > > > > > > > > > > > > > Engine > > > > > > > > > > > > > > > and oVirt Node require different versions of RHEL or > > > > > > > > > > > CentOS. > > > > > > > > > > > > > > > > > > > > > > > > > > > > I read that the Engine for oVirt 4.4.0 requires RHEL > > > > > > > > > > > (or > > > > > > > > > > > > > > > > CentOS) > > > > > > > > > > > > > > > 8.2, whereas each Node requires 7.x (although I'll plan > > > > > > > > > > > to just > > > > > > > > > > > > > > > > use the > > > > > > > > > > > > > > > oVirt Node ISO). > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > I'm also wondering about storage. > > > > > > > > > > > I don't really like the idea of using local storage, > > > > > > > > > > > but a > > > > > > > > > > > > > > > > single > > > > > > > > > > > > > > > NFS server would also be a single point of failure, and > > > > > > > > > > > Gluster > > > > > > > > > > > > > > > > would > > > > > > > > > > > > > > > be too expensive to deploy, so at this point, I'm > > > > > > > > > > > leaning > > > > > > > > > > > > > > > > towards using > > > > > > > > > > > > > > > local storage. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Any advice or clarity would be greatly appreciated. > > > > > > > > > > > > > > > Thanks, > > > > > > > > > > > David > > > > > > > > > > > > > > > Sent with ProtonMail Secure Email. > > > > > > > > > > > > > > > Users mailing list -- users@ovirt.org > > > > > > > > > > > To unsubscribe send an email to users-le...@ovirt.org > > > > > > > > > > > Privacy Statement: > > > > > > > > > > > https://www.ovirt.org/privacy-policy.html > > > > > > > > > > > > > > > > oVirt Code of Conduct: > > > > > > > https://www.ovirt.org/community/about/community-guidelines/ > > > > > > > > > > > > > > > > > > > > > > > > > > > List Archives: > > > > > > > > > > > > > > > > > > > > https://lists.ovirt.org/archives/list/users@ovirt.org/message/RGHCN356DXJEDR5FJ7SXSBHBF5FYRWIN/ > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Users mailing list -- users@ovirt.org > > > > > > > > To unsubscribe send an email to users-le...@ovirt.org > > > > > > > > Privacy Statement: https://www.ovirt.org/privacy-policy.html > > > > > > > > oVirt Code of Conduct: > > > > > > > > https://www.ovirt.org/community/about/community-guidelines/ > > > > > > > > List Archives: > > > > > > > > https://lists.ovirt.org/archives/list/users@ovirt.org/message/TF63JRHWDBCJWDRW66I3NF6WUKKFHNCX/ > > > > > > > > > > > > > > > > > > > > > Users mailing list -- users@ovirt.org > > > > > > To unsubscribe send an email to users-le...@ovirt.org > > > > > > Privacy Statement: https://www.ovirt.org/privacy-policy.html > > > > > > oVirt Code of Conduct: > > > > > > https://www.ovirt.org/community/about/community-guidelines/ > > > > > > > > > > > List Archives: > > > > > > > https://lists.ovirt.org/archives/list/users@ovirt.org/message/GKCVKWJJ56ITACRPSMZGCE2Z47JWUZQI/ > > > > > > > > > > > > > > > > > > > > Users mailing list -- users@ovirt.org > > > > > To unsubscribe send an email to users-le...@ovirt.org > > > > > Privacy Statement: https://www.ovirt.org/privacy-policy.html > > > > > oVirt Code of Conduct: > > > > > https://www.ovirt.org/community/about/community-guidelines/ > > > > > > > > > > List Archives: > > > > > > https://lists.ovirt.org/archives/list/users@ovirt.org/message/JYFSHSV4IZBELAUA6A5M6TINGNBWBP7G/ > > > > > Users mailing list -- users@ovirt.org > > To unsubscribe send an email to users-le...@ovirt.org > > Privacy Statement: https://www.ovirt.org/privacy-policy.html > > oVirt Code of Conduct: > https://www.ovirt.org/community/about/community-guidelines/ > > List Archives: > https://lists.ovirt.org/archives/list/users@ovirt.org/message/22BOW47SBSCDLHPLZFYGJSJLZOBJ6WRA/ > > _______________________________________________ > Users mailing list -- users@ovirt.org > To unsubscribe send an email to users-le...@ovirt.org > Privacy Statement: https://www.ovirt.org/privacy-policy.html > oVirt Code of Conduct: > https://www.ovirt.org/community/about/community-guidelines/ > List Archives: > https://lists.ovirt.org/archives/list/users@ovirt.org/message/HPFH7QJA6WSCBTD7LHSO2HXAAZMWDODX/ >
_______________________________________________ Users mailing list -- users@ovirt.org To unsubscribe send an email to users-le...@ovirt.org Privacy Statement: https://www.ovirt.org/privacy-policy.html oVirt Code of Conduct: https://www.ovirt.org/community/about/community-guidelines/ List Archives: https://lists.ovirt.org/archives/list/users@ovirt.org/message/JIIW6Z6T7TBEVQKDV5OQQGCOPDVWVWKB/