Local storage & shared in same cluster
Hi, Is there a technical reason for not being able to use the local host storage if that host is already in a cluster with shared storage? For instance, if a user has a server with with a lot of storage and he wants to use that storage for non HA VMs that will only run over that host while, that host is also part of a cluster running VMs in a shared storage. If there's no technical reason/limitation behind this and this is a design choice, do you consider this to be in a roadmap? Thanks! Sebastián Greco IT Consultant Cloud Computing - Red Hat - VMware - Zimbra www.essiprojects.com *www.essiprojects.co.uk <http://www.essiprojects.co.uk>* Pl. Prim, 4-5 Pral 2a · T:+34 977 221 182 · M: +34 619 985 161 F: +34 977 230 170 · 43001 Tarragona Spain 120 Pall Mall · T:+44 207 101 0778 · F: +44 843 538 3112 · SW1Y 5ED *London* UK
By design, host local storage is accessible from the host itself only and that's why it cannot be used in shared storage Data Centres. On Thu, Oct 27, 2016 at 9:59 AM, Sebastian Greco <sgreco@essiprojects.com> wrote:
Hi,
Is there a technical reason for not being able to use the local host storage if that host is already in a cluster with shared storage?
For instance, if a user has a server with with a lot of storage and he wants to use that storage for non HA VMs that will only run over that host while, that host is also part of a cluster running VMs in a shared storage.
If there's no technical reason/limitation behind this and this is a design choice, do you consider this to be in a roadmap?
Thanks!
Sebastián Greco IT Consultant Cloud Computing - Red Hat - VMware - Zimbra www.essiprojects.com *www.essiprojects.co.uk <http://www.essiprojects.co.uk>*
Pl. Prim, 4-5 Pral 2a · T:+34 977 221 182 · M: +34 619 985 161 F: +34 977 230 170 · 43001 Tarragona Spain 120 Pall Mall · T:+44 207 101 0778 · F: +44 843 538 3112 · SW1Y 5ED *London* UK
_______________________________________________ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
Ohh thx for the answer :) That's why the running VMs on that particular storage would only run on that host. One can do that in other hypervisors like vsphere or hyper-v. It would be a nice thing to have in rhv too imho. On Thu, Oct 27, 2016 at 12:44 PM, Elad Ben Aharon <ebenahar@redhat.com> wrote:
By design, host local storage is accessible from the host itself only and that's why it cannot be used in shared storage Data Centres.
On Thu, Oct 27, 2016 at 9:59 AM, Sebastian Greco <sgreco@essiprojects.com> wrote:
Hi,
Is there a technical reason for not being able to use the local host storage if that host is already in a cluster with shared storage?
For instance, if a user has a server with with a lot of storage and he wants to use that storage for non HA VMs that will only run over that host while, that host is also part of a cluster running VMs in a shared storage.
If there's no technical reason/limitation behind this and this is a design choice, do you consider this to be in a roadmap?
Thanks!
Sebastián Greco IT Consultant Cloud Computing - Red Hat - VMware - Zimbra www.essiprojects.com *www.essiprojects.co.uk <http://www.essiprojects.co.uk>*
Pl. Prim, 4-5 Pral 2a · T:+34 977 221 182 · M: +34 619 985 161 F: +34 977 230 170 · 43001 Tarragona Spain 120 Pall Mall · T:+44 207 101 0778 · F: +44 843 538 3112 · SW1Y 5ED *London* UK
_______________________________________________ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
On Thu, Oct 27, 2016 at 1:58 PM, Sebastian Greco <sgreco@essiprojects.com> wrote:
Ohh thx for the answer :) That's why the running VMs on that particular storage would only run on that host. One can do that in other hypervisors like vsphere or hyper-v. It would be a nice thing to have in rhv too imho.
Asked many times... Latest relevant thread just before 4.0 release with some technical details here: http://lists.ovirt.org/pipermail/users/2016-May/039772.html
On 27 October 2016 at 14:58, Sebastian Greco <sgreco@essiprojects.com> wrote:
Ohh thx for the answer :) That's why the running VMs on that particular storage would only run on that host. One can do that in other hypervisors like vsphere or hyper-v. It would be a nice thing to have in rhv too imho.
Well, you can partially do this, with the scratch-pad hook, if you don't care about the local data going away when the VM shuts down. https://github.com/oVirt/vdsm/tree/master/vdsm_hooks/scratchpad We've been successfully using this in ovirt CI to keep intermediate build process I/O out of our central storage. It seems to me that short-lived non-migrating but stateful VMs are not very interesting as a use case for RHV customers. When you run you databases and applications on RHV, it seems to me you typically want your data to stick around and your VMs to survive a hypervisor crash. Then again, I do not have any data to substantiate or negate this claim. -- Barak Korren bkorren@redhat.com RHEV-CI Team
On Sun, Oct 30, 2016 at 10:42 AM, Barak Korren <bkorren@redhat.com> wrote:
VMs are not very interesting as a use case for RHV customers. When y
Thx for the answsers. I see that it's the second time that someone from RH points out that customers are not interested in this feature. While I can't argue with that, what I do can say is that "non-customers" (most of companies out there using vsphere or hyper-v) feel dissapointed towards this solution for things like this one (for this case, 2 of my customers are missing this, we are deploying RHV to one of them this week). I don't see how this lack of flexibility is something good, and so far from my experience with customers which I'm trying to convince to start using RHV, when they finally do agree to start with one or two servers (following the RHCI roadmap evolution to the hybrid cloud), they see things like this and dismiss this solution sooner than later. Anyways, question has been answer "yes, is technically possible but by design it is not going to happen", and I wouldn't like to convert this thread or abuse your kindness deviating the subject :) Thx again guys for the help,
On Mon, Oct 31, 2016 at 9:28 AM, Sebastian Greco <sgreco@essiprojects.com> wrote:
On Sun, Oct 30, 2016 at 10:42 AM, Barak Korren <bkorren@redhat.com> wrote:
VMs are not very interesting as a use case for RHV customers. When y
Thx for the answsers. I see that it's the second time that someone from RH points out that customers are not interested in this feature. While I can't argue with that, what I do can say is that "non-customers" (most of companies out there using vsphere or hyper-v) feel dissapointed towards this solution for things like this one (for this case, 2 of my customers are missing this, we are deploying RHV to one of them this week).
I don't see how this lack of flexibility is something good, and so far from my experience with customers which I'm trying to convince to start using RHV, when they finally do agree to start with one or two servers (following the RHCI roadmap evolution to the hybrid cloud), they see things like this and dismiss this solution sooner than later.
Anyways, question has been answer "yes, is technically possible but by design it is not going to happen",
Not sure how you got this impression from current discussion. On the contrary - people are working on this, and it was planned to be completed in 4.0, but eventually postponed. Best,
and I wouldn't like to convert this thread or abuse your kindness deviating the subject :)
Thx again guys for the help,
_______________________________________________ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
-- Didi
Maybe I missread, sorry. I interpreted that by design this was not a feature and that it was not in any roadmap as current customers don't need it, or at least I understood that from previous posts. Happy to know this is planned to be a feature! :) Again guys, thanks a lot for the help! Sebastián Greco IT Consultant Cloud Computing - Red Hat - VMware - Zimbra www.essiprojects.com *www.essiprojects.co.uk <http://www.essiprojects.co.uk>* Pl. Prim, 4-5 Pral 2a · T:+34 977 221 182 · M: +34 619 985 161 F: +34 977 230 170 · 43001 Tarragona Spain 120 Pall Mall · T:+44 207 101 0778 · F: +44 843 538 3112 · SW1Y 5ED *London* UK On Mon, Oct 31, 2016 at 8:59 AM, Yedidyah Bar David <didi@redhat.com> wrote:
On Mon, Oct 31, 2016 at 9:28 AM, Sebastian Greco <sgreco@essiprojects.com> wrote:
On Sun, Oct 30, 2016 at 10:42 AM, Barak Korren <bkorren@redhat.com>
VMs are not very interesting as a use case for RHV customers. When y
Thx for the answsers. I see that it's the second time that someone from RH points out that customers are not interested in this feature. While I can't argue with that, what I do can say is that "non-customers" (most of companies out there using vsphere or hyper-v) feel dissapointed towards
solution for things like this one (for this case, 2 of my customers are missing this, we are deploying RHV to one of them this week).
I don't see how this lack of flexibility is something good, and so far from my experience with customers which I'm trying to convince to start using RHV, when they finally do agree to start with one or two servers (following the RHCI roadmap evolution to the hybrid cloud), they see things like
wrote: this this
and dismiss this solution sooner than later.
Anyways, question has been answer "yes, is technically possible but by design it is not going to happen",
Not sure how you got this impression from current discussion. On the contrary - people are working on this, and it was planned to be completed in 4.0, but eventually postponed.
Best,
and I wouldn't like to convert this thread or abuse your kindness deviating the subject :)
Thx again guys for the help,
_______________________________________________ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
-- Didi
--351cHtD1l1mj1p9igIErMfxngtEjXVIsG Content-Type: multipart/mixed; boundary="9Tv7cbMStURuEiRbhesQriiguFIKRalvu"; protected-headers="v1" From: Sven Kieske <s.kieske@mittwald.de> To: users@ovirt.org Message-ID: <a736e6ed-e112-1777-2cc6-cfff18feebb4@mittwald.de> Subject: Re: [ovirt-users] Local storage & shared in same cluster References: <CABLLjPrATCACHtoe5Eg2KweXuUxSfccPb5zO9hZ3kvMAhiXPKQ@mail.gmail.com> <CAOpfNcUvN_D+prLaZgueKanxKXb7W1gCrbQVAoH6X9CRcP2k-w@mail.gmail.com> <CABLLjPqqDds5HvzkL_FAjX0MA3ooHHHmTYVa3ZBjD90t7JHBcg@mail.gmail.com> <CAGJrMmoYTT98at=EJ+HOypP-M-1-KV0szyJiEMXKwxRShL7dFg@mail.gmail.com> <CABLLjPpr8npzRsQ5KhC61kVQ52QKQcfkBRcjS_AC5LtE=WAZ9g@mail.gmail.com> <CAHRwYXtKemik-ZaZgyg=koNLm7577hOQ_ZHNJZ_1nKE_5tj25Q@mail.gmail.com> In-Reply-To: <CAHRwYXtKemik-ZaZgyg=koNLm7577hOQ_ZHNJZ_1nKE_5tj25Q@mail.gmail.com> --9Tv7cbMStURuEiRbhesQriiguFIKRalvu Content-Type: text/plain; charset=windows-1252 Content-Transfer-Encoding: quoted-printable On 31/10/16 08:59, Yedidyah Bar David wrote:
Not sure how you got this impression from current discussion. On the contrary - people are working on this, and it was planned to be complet= ed in 4.0, but eventually postponed.
This is really interesting. But don't you need to remove the master storage domain concept for this, plus spm? The RFE for this was closed as "wontfix": https://bugzilla.redhat.com/show_bug.cgi?id=3D1080372 Am I mixing things up? It would be nice if someone could clarify this. --=20 Mit freundlichen Gr=FC=DFen / Regards Sven Kieske Systemadministrator Mittwald CM Service GmbH & Co. KG K=F6nigsberger Stra=DFe 6 32339 Espelkamp T: +495772 293100 F: +495772 293333 https://www.mittwald.de Gesch=E4ftsf=FChrer: Robert Meyer St.Nr.: 331/5721/1033, USt-IdNr.: DE814773217, HRA 6640, AG Bad Oeynhause= n Komplement=E4rin: Robert Meyer Verwaltungs GmbH, HRB 13260, AG Bad Oeynha= usen --9Tv7cbMStURuEiRbhesQriiguFIKRalvu-- --351cHtD1l1mj1p9igIErMfxngtEjXVIsG Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.22 (GNU/Linux) iQIcBAEBAgAGBQJYGeyuAAoJEMby9TMDAbQRWlAP/0ZVSz8kqw4gRDw9MYi61UqP vaAWJBsz/tHXW8mD7Gb7RebGOnCl7oRGBDn5xyz4+zc5VrroTbl+FRTX3pvG+5F3 smjaWE8tqzCFfyDbB4t8Hva44VZ6fPB3KUiTz5zfM17MDkxJAjieN3GWlX6WthrA KIzIqjMKQntkpQlGj28HR0JKCFcavir1APfkPUCycRyVnjgJaRt59Gk3E5b4iuco J1rW7N/STwCF8Hu78tTFZ1zY2fTB1pF1tqTzHOygawDmB2wSIBAToRYBJ3IEEs8T eYivyauMI5lagIgwmPt99BLRiel4NIAmWuzZNxSHug7AKTKP9MyP4k81tYlnddWc wapCkTNhTLcVnGEMpmqWUPe5eak8g4Btgdctg6wHesC0qFIXyFcK8c/8eHHjZPl4 J9j9n//ZqEDW5dfHG8i4WQpACI+Fx4Ev46f0Eo10SXnj3rJmnX3udh88NGeq4TWt ohU2eO2g38ZslHGboKx20NGYxJ22lAfheaABjI38o06eOmcIdQVVQMRgMYADXMuX 8pFaI6gyBScG50U5sQ7d0eDeOvp2B3tZK2VcPYUxwwkzktj1okrGoJbBQbIauqHa u/5FD1hRiCRTLvAjjhXbEokwiedopQ20LvlNKAor0X4Mu0UnGxyjmOxe0ojz6U9P 7zqoLvXmhPEJs3IT3Uma =slpX -----END PGP SIGNATURE----- --351cHtD1l1mj1p9igIErMfxngtEjXVIsG--
On 31 October 2016 at 09:28, Sebastian Greco <sgreco@essiprojects.com> wrote:
On Sun, Oct 30, 2016 at 10:42 AM, Barak Korren <bkorren@redhat.com> wrote:
VMs are not very interesting as a use case for RHV customers. When y
Thx for the answsers. I see that it's the second time that someone from RH points out that customers are not interested in this feature. While I can't argue with that, what I do can say is that "non-customers" (most of companies out there using vsphere or hyper-v) feel dissapointed towards this solution for things like this one (for this case, 2 of my customers are missing this, we are deploying RHV to one of them this week).
Please don't take my statement as being official in any way. Despite writing from a @redhat.com address, I'm writing my personal thoughts. I have stated that I've no data to back what I've said. This is all just a guess based on what I know of oVirt/RHV development processes. I'm most certainly not someone who makes decisions about any of theses things.
I don't see how this lack of flexibility is something good, and so far from my experience with customers which I'm trying to convince to start using RHV, when they finally do agree to start with one or two servers (following the RHCI roadmap evolution to the hybrid cloud), they see things like this and dismiss this solution sooner than later.
Please do not take my statement as indicating of any conscious design decision. I was just trying to gauge where oVirt/RHV development might head given that RedHat typically puts its resources where its current and potential customers tell it do. Case to point: 1. Ephemeral local state VMs are supported with the scrathcpad hook because its been shown to be useful for Build/Test/CI systems. 2. Singular host with local storage and non-migrating VMs is supported for cases where one simply wants resource convergence. The 3rd case we're discussing here where the same host can run both local persistent VMs and migrating ones had not been supported so far. I'm __guessing__ that this is because demand seen so far did not outweigh the technical difficulty to achieve this (Just to indicate the difficulty, the SPM was planned to be removed in 4.0, it did not make it).
Anyways, question has been answer "yes, is technically possible but by design it is not going to happen", and I wouldn't like to convert this thread or abuse your kindness deviating the subject :)
This is definitely not the bottom line, I way trying to guess and explain why this __did_not__ happen __so_far__. I never meant to say it will not. If you are a RHV reseller or integrator, your input is very valuable for RedHat. While this list is one way to reach some RedHat developers, you should certainly make an effort to use other channels available to you to make your input known. -- Barak Korren bkorren@redhat.com RHEV-CI Team
Thx for the advice Barak! I was looking for a roadmap to follow, like the "trello" one used for the openshift project https://trello.com/b/nlLwlKoz/atomicopenshift-roadmap but I was not able to find one. I love to hear things like the "SPM was planned to be removed" as it gives me a better understanding of the direction the project has. Any good resource to follow this and other things to come? Thanks again for the help! Sebastián Greco IT Consultant Cloud Computing - Red Hat - VMware - Zimbra www.essiprojects.com *www.essiprojects.co.uk <http://www.essiprojects.co.uk>* Pl. Prim, 4-5 Pral 2a · T:+34 977 221 182 · M: +34 619 985 161 F: +34 977 230 170 · 43001 Tarragona Spain 120 Pall Mall · T:+44 207 101 0778 · F: +44 843 538 3112 · SW1Y 5ED *London* UK On Mon, Oct 31, 2016 at 9:36 AM, Barak Korren <bkorren@redhat.com> wrote:
On 31 October 2016 at 09:28, Sebastian Greco <sgreco@essiprojects.com> wrote:
On Sun, Oct 30, 2016 at 10:42 AM, Barak Korren <bkorren@redhat.com>
VMs are not very interesting as a use case for RHV customers. When y
Thx for the answsers. I see that it's the second time that someone from RH points out that customers are not interested in this feature. While I can't argue with that, what I do can say is that "non-customers" (most of companies out there using vsphere or hyper-v) feel dissapointed towards
wrote: this
solution for things like this one (for this case, 2 of my customers are missing this, we are deploying RHV to one of them this week).
Please don't take my statement as being official in any way. Despite writing from a @redhat.com address, I'm writing my personal thoughts.
I have stated that I've no data to back what I've said. This is all just a guess based on what I know of oVirt/RHV development processes. I'm most certainly not someone who makes decisions about any of theses things.
I don't see how this lack of flexibility is something good, and so far from my experience with customers which I'm trying to convince to start using RHV, when they finally do agree to start with one or two servers (following the RHCI roadmap evolution to the hybrid cloud), they see things like this and dismiss this solution sooner than later.
Please do not take my statement as indicating of any conscious design decision. I was just trying to gauge where oVirt/RHV development might head given that RedHat typically puts its resources where its current and potential customers tell it do. Case to point:
1. Ephemeral local state VMs are supported with the scrathcpad hook because its been shown to be useful for Build/Test/CI systems. 2. Singular host with local storage and non-migrating VMs is supported for cases where one simply wants resource convergence.
The 3rd case we're discussing here where the same host can run both local persistent VMs and migrating ones had not been supported so far. I'm __guessing__ that this is because demand seen so far did not outweigh the technical difficulty to achieve this (Just to indicate the difficulty, the SPM was planned to be removed in 4.0, it did not make it).
Anyways, question has been answer "yes, is technically possible but by design it is not going to happen", and I wouldn't like to convert this thread or abuse your kindness deviating the subject :)
This is definitely not the bottom line, I way trying to guess and explain why this __did_not__ happen __so_far__. I never meant to say it will not.
If you are a RHV reseller or integrator, your input is very valuable for RedHat. While this list is one way to reach some RedHat developers, you should certainly make an effort to use other channels available to you to make your input known.
-- Barak Korren bkorren@redhat.com RHEV-CI Team
participants (6)
-
Barak Korren -
Elad Ben Aharon -
Gianluca Cecchi -
Sebastian Greco -
Sven Kieske -
Yedidyah Bar David