oVirt: single host install

Use case: small sites with a minimum number of vm's. Is there such a thing as a single host install? Is it valid for production use? What kind of storage?

On Mon, May 21, 2018 at 7:49 AM, <ovirt@fateknollogee.com> wrote:
Use case: small sites with a minimum number of vm's.
Is there such a thing as a single host install?
In the past we had the all-in-one mode but we deprecated it. Now the suggested mode is hosted-engine since you could expand it adding other hosts in the future.
Is it valid for production use?
With a single host the upgrades will become more intrusive: without the capability to migrate your VMs on other hosts at upgrade time, you will be required to bring down everything.
What kind of storage?
NFS in loopback could be problematic, I'd suggest gluster in replica 1 or iSCSI.
_______________________________________________ Users mailing list -- users@ovirt.org To unsubscribe send an email to users-leave@ovirt.org

Hi, Simone Tiraboschi <stirabos@redhat.com> writes:
On Mon, May 21, 2018 at 7:49 AM, <ovirt@fateknollogee.com> wrote:
Use case: small sites with a minimum number of vm's.
Is there such a thing as a single host install?
In the past we had the all-in-one mode but we deprecated it. Now the suggested mode is hosted-engine since you could expand it adding other hosts in the future.
I am running in this configuration and have had little problem. I migrated from an old vmware-server platform, and, modulo a few hiccups along the way and a few false starts as I was installing ovirt, it's been pretty stable for me!
Is it valid for production use?
With a single host the upgrades will become more intrusive: without the capability to migrate your VMs on other hosts at upgrade time, you will be required to bring down everything.
This is true -- I have to bring everything down when I want to upgrade the system, especially the host itself. So I don't upgrade as often as I might if I had multiple hosts where I could migrate.
What kind of storage?
NFS in loopback could be problematic, I'd suggest gluster in replica 1 or iSCSI.
I'm running loopback NFS and I've not encountered any issues. I've been running this way since 2016-10-22. I did not understand Gluster enough and wasn't sure how I could make a "replica 1" -- everything seemed to imply you *NEEDED* 3 gluster hosts. So I went with what I knew -- NFS. This might be problematic if I move forward to a multi-host platform as I'll have to "migrate" my storage -- specifically for hosted-engine -- which IIRC requires a re-install (or some other drastic measure). -derek -- Derek Atkins, SB '93 MIT EE, SM '95 MIT Media Laboratory Member, MIT Student Information Processing Board (SIPB) URL: http://web.mit.edu/warlord/ PP-ASEL-IA N1NWH warlord@MIT.EDU PGP key available

-----Original Message----- From: Derek Atkins [mailto:warlord@MIT.EDU] Sent: Monday, May 21, 2018 8:33 AM To: Simone Tiraboschi <stirabos@redhat.com> Cc: users <users@ovirt.org>; ovirt@fateknollogee.com Subject: [ovirt-users] Re: oVirt: single host install Hi, Simone Tiraboschi <stirabos@redhat.com> writes:
On Mon, May 21, 2018 at 7:49 AM, <ovirt@fateknollogee.com> wrote:
Use case: small sites with a minimum number of vm's.
Is there such a thing as a single host install?
In the past we had the all-in-one mode but we deprecated it. Now the suggested mode is hosted-engine since you could expand it adding other hosts in the future.
Sounds like "local storage" is useless then? I am running in this configuration and have had little problem. I migrated from an old vmware-server platform, and, modulo a few hiccups along the way and a few false starts as I was installing ovirt, it's been pretty stable for me! Did you use NFS for all storage domains?
Is it valid for production use?
With a single host the upgrades will become more intrusive: without the capability to migrate your VMs on other hosts at upgrade time, you will be required to bring down everything.
This is true -- I have to bring everything down when I want to upgrade the system, especially the host itself. So I don't upgrade as often as I might if I had multiple hosts where I could migrate. How can you do it without a second host and importing with a temporary storage domain?
What kind of storage?
NFS in loopback could be problematic, I'd suggest gluster in replica 1 or iSCSI.
I'm running loopback NFS and I've not encountered any issues. I've been running this way since 2016-10-22. I did not understand Gluster enough and wasn't sure how I could make a "replica 1" -- everything seemed to imply you *NEEDED* 3 gluster hosts. So I went with what I knew -- NFS. This might be problematic if I move forward to a multi-host platform as I'll have to "migrate" my storage -- specifically for hosted-engine -- which IIRC requires a re-install (or some other drastic measure). -derek -- Derek Atkins, SB '93 MIT EE, SM '95 MIT Media Laboratory Member, MIT Student Information Processing Board (SIPB) URL: https://urldefense.proofpoint.com/v2/url?u=http-3A__web.mit.edu_warlord_&d=DwIGaQ&c=Vxt5e0Osvvt2gflwSlsJ5DmPGcPvTRKLJyp031rXjhg&r=FiPhL0Cl1ymZlnTyAIIL75tE4L0reHcDdD-7wUtUGHA&m=aceiPp0tdZww5F1VjIbEi55JnMpkWzYlKSpsa56bcG8&s=Sxh76egRoJB5MNDdFRR9NubIiPK1tqZlSyW9tt4XzRQ&e= PP-ASEL-IA N1NWH warlord@MIT.EDU PGP key available _______________________________________________ Users mailing list -- users@ovirt.org To unsubscribe send an email to users-leave@ovirt.org

Hi, Justin Zygmont <jzygmont@proofpoint.com> writes:
-----Original Message----- From: Derek Atkins [mailto:warlord@MIT.EDU] Sent: Monday, May 21, 2018 8:33 AM To: Simone Tiraboschi <stirabos@redhat.com> Cc: users <users@ovirt.org>; ovirt@fateknollogee.com Subject: [ovirt-users] Re: oVirt: single host install
Hi,
Simone Tiraboschi <stirabos@redhat.com> writes:
On Mon, May 21, 2018 at 7:49 AM, <ovirt@fateknollogee.com> wrote:
Use case: small sites with a minimum number of vm's.
Is there such a thing as a single host install?
In the past we had the all-in-one mode but we deprecated it. Now the suggested mode is hosted-engine since you could expand it adding other hosts in the future.
Sounds like "local storage" is useless then?
IMHO, yes. When I installed (4.0) you could not use it for hosted-engine storage. Don't know if that changed, but what's the point of using different storage methods? I already had to set up a local NFS (pr theoretically Gluster) for HE -- so might as well re-use that for my main storage too!
I am running in this configuration and have had little problem. I migrated from an old vmware-server platform, and, modulo a few hiccups along the way and a few false starts as I was installing ovirt, it's been pretty stable for me!
Did you use NFS for all storage domains?
Yes. Both HE and Main Data storage domains are backed by SSD. My ISO domain is backed by spinning rust.
Is it valid for production use?
With a single host the upgrades will become more intrusive: without the capability to migrate your VMs on other hosts at upgrade time, you will be required to bring down everything.
This is true -- I have to bring everything down when I want to upgrade the system, especially the host itself. So I don't upgrade as often as I might if I had multiple hosts where I could migrate.
How can you do it without a second host and importing with a temporary storage domain?
I just shut everything down. So long as it's planned my users can handle a 30-60 minute outage. And this is only when I update the host. I can update the Engine on its own, and often will update my VMs simultaneously to minimize downtime. But I'm okay with some downtime. -derek PS: I don't know what mailer you used, but I had a very hard time differentiating your responses from mine. Hopefully I did not miss one of your questions. -- Derek Atkins, SB '93 MIT EE, SM '95 MIT Media Laboratory Member, MIT Student Information Processing Board (SIPB) URL: http://web.mit.edu/warlord/ PP-ASEL-IA N1NWH warlord@MIT.EDU PGP key available

This is true -- I have to bring everything down when I want to upgrade the system, especially the host itself. So I don't upgrade as often as I might if I had multiple hosts where I could migrate.
How can you do it without a second host and importing with a temporary storage domain?
I just shut everything down. So long as it's planned my users can >handle a 30-60 minute outage. And this is only when I update the >host. I can update the Engine on its own, and often will update my VMs >simultaneously to minimize downtime. But I'm okay with some >downtime.
I see, so you just did an in place update of engine I guess, what if you want to update the node as well, and install a new HE? You'd lose the locally stored NFS domains. Or what if the engine update stuffs up, there'd be no way to access the admin portal right?

Hi, Justin Zygmont <jzygmont@proofpoint.com> writes:
This is true -- I have to bring everything down when I want to upgrade the system, especially the host itself. So I don't upgrade as often as I might if I had multiple hosts where I could migrate.
How can you do it without a second host and importing with a temporary storage domain?
I just shut everything down. So long as it's planned my users can >handle a 30-60 minute outage. And this is only when I update the >host. I can update the Engine on its own, and often will update my VMs >simultaneously to minimize downtime. But I'm okay with some >downtime.
I see, so you just did an in place update of engine I guess, what if you want to update the node as well, and install a new HE? You'd lose the locally stored NFS domains. Or what if the engine update stuffs up, there'd be no way to access the admin portal right?
Yes, log into engine and run: yum update 'ovirt-*-setup' Then engine-setup. As for the host, I'm not using Node. It's just a regular CentOS system with the ovirt host software installed. I update it the normal way you'd update any other CentOS system: yum update I don't understand what you mean by "install a new HE"? If this is a single-host system, what do you mean by installing a new hosted engine? If you're asking about starting with one host and adding a second host, then that returns back to the previous statement that you'd have to migrate HE to a new (Gluster-based) storage system. If you stayed with NFS, then you "master host" could never go down or your HE would also go down (because its [NFS-based] storage would go away). You're right that if the Engine update breaks somehow then the admin portal would go away -- but I'm never updating systems through the portal. It's always via SSH (or worst case local console on the host). If the engine update breaks so much that even ssh wont work, then I'm definitely in trouble. That's what backups are for! :) -derek -- Derek Atkins, SB '93 MIT EE, SM '95 MIT Media Laboratory Member, MIT Student Information Processing Board (SIPB) URL: http://web.mit.edu/warlord/ PP-ASEL-IA N1NWH warlord@MIT.EDU PGP key available

On Tue, May 22, 2018 at 12:17 AM, Justin Zygmont <jzygmont@proofpoint.com> wrote:
-----Original Message----- From: Derek Atkins [mailto:warlord@MIT.EDU] Sent: Monday, May 21, 2018 8:33 AM To: Simone Tiraboschi <stirabos@redhat.com> Cc: users <users@ovirt.org>; ovirt@fateknollogee.com Subject: [ovirt-users] Re: oVirt: single host install
Hi,
Simone Tiraboschi <stirabos@redhat.com> writes:
On Mon, May 21, 2018 at 7:49 AM, <ovirt@fateknollogee.com> wrote:
Use case: small sites with a minimum number of vm's.
Is there such a thing as a single host install?
In the past we had the all-in-one mode but we deprecated it. Now the suggested mode is hosted-engine since you could expand it adding other hosts in the future.
Sounds like "local storage" is useless then?
I am running in this configuration and have had little problem. I migrated from an old vmware-server platform, and, modulo a few hiccups along the way and a few false starts as I was installing ovirt, it's been pretty stable for me!
Did you use NFS for all storage domains?
Is it valid for production use?
With a single host the upgrades will become more intrusive: without the capability to migrate your VMs on other hosts at upgrade time, you will be required to bring down everything.
This is true -- I have to bring everything down when I want to upgrade the system, especially the host itself. So I don't upgrade as often as I might if I had multiple hosts where I could migrate.
How can you do it without a second host and importing with a temporary storage domain?
What kind of storage?
NFS in loopback could be problematic, I'd suggest gluster in replica 1 or iSCSI.
I'm running loopback NFS and I've not encountered any issues. I've been running this way since 2016-10-22. I did not understand Gluster enough and wasn't sure how I could make a "replica 1" -- everything seemed to imply you *NEEDED* 3 gluster hosts. So I went with what I knew -- NFS.
We did add support for single node gluster volume in 4.2 - see https://www.ovirt.org/documentation/gluster-hyperconverged/chap-Single_node_...
This might be problematic if I move forward to a multi-host platform as I'll have to "migrate" my storage -- specifically for hosted-engine -- which IIRC requires a re-install (or some other drastic measure).
-derek
-- Derek Atkins, SB '93 MIT EE, SM '95 MIT Media Laboratory Member, MIT Student Information Processing Board (SIPB) URL: https://urldefense.proofpoint.com/v2/url?u=http-3A__web.mit. edu_warlord_&d=DwIGaQ&c=Vxt5e0Osvvt2gflwSlsJ5DmPGcPvTRKLJyp031rXjhg&r= FiPhL0Cl1ymZlnTyAIIL75tE4L0reHcDdD-7wUtUGHA&m= aceiPp0tdZww5F1VjIbEi55JnMpkWzYlKSpsa56bcG8&s= Sxh76egRoJB5MNDdFRR9NubIiPK1tqZlSyW9tt4XzRQ&e= PP-ASEL-IA N1NWH warlord@MIT.EDU PGP key available _______________________________________________ Users mailing list -- users@ovirt.org To unsubscribe send an email to users-leave@ovirt.org _______________________________________________ Users mailing list -- users@ovirt.org To unsubscribe send an email to users-leave@ovirt.org

Hi, Sahina Bose <sabose@redhat.com> writes:
I'm running loopback NFS and I've not encountered any issues. I've been running this way since 2016-10-22. I did not understand Gluster enough and wasn't sure how I could make a "replica 1" -- everything seemed to imply you *NEEDED* 3 gluster hosts. So I went with what I knew -- NFS.
We did add support for single node gluster volume in 4.2 - see https://www.ovirt.org/documentation/gluster-hyperconverged/chap-Single_node_...
Good to know, and thank you for the link. I started with 4.0, which explains why I did not go this route. I've upgraded along the way, but the upgrades wont let me easily change out the underlying storage mechanisms without fully reinstalling the HE. Still, next time I do a system refresh I'll definitely consider this. Thanks, -derek -- Derek Atkins, SB '93 MIT EE, SM '95 MIT Media Laboratory Member, MIT Student Information Processing Board (SIPB) URL: http://web.mit.edu/warlord/ PP-ASEL-IA N1NWH warlord@MIT.EDU PGP key available
participants (5)
-
Derek Atkins
-
Justin Zygmont
-
ovirt@fateknollogee.com
-
Sahina Bose
-
Simone Tiraboschi