Re: [pve-devel] LXC volumes on ZFS over iSCSI

On Wed, 12 Oct 2016 05:41:26 +0200 (CEST) Alexandre DERUMIER wrote: > > Well, this is very dependent of nfs server implementation quality. > I'm running vm on netapp san though nfs 4.1, and I have very good performance. > netapp is a purpose build NFS based storage

Re: [pve-devel] LXC volumes on ZFS over iSCSI

>>And what about CIFS? I can't comment about CIFS, I never run vm on it. But what I known , is that it's really depending of CIFS version implementation, both client && server. Old CIFS version (smb 1.x) (windows2003/xp), was a very bad protocol (very chatty). I known that last windows2012

Re: [pve-devel] LXC volumes on ZFS over iSCSI

Alexandre Derumier Ingénieur système et stockage Manager Infrastructure Fixe : +33 3 59 82 20 10 125 Avenue de la république 59110 La Madeleine [ https://twitter.com/OdisoHosting ] [ https://twitter.com/mindbaz ] [ https://www.linkedin.com/company/odiso ] [

Re: [pve-devel] LXC volumes on ZFS over iSCSI

> I'm curious to see result of: > >>It seems to work :-) Ok, so it shouldn't be too diffcult to add|remove lun in activate|deactiveate volume :) About NFS vs Iscsi, I think we should add both if we can. Because not everybody love both NFS && ISCSI. (I'm already seeing the long debate in

Re: [pve-devel] LXC volumes on ZFS over iSCSI

> >>NFS does not provide good iops and has the ability to bring down a > >>node. IMHO NFS is only useful as filestorage server for backups and > >>iso images. > > Well, this is very dependent of nfs server implementation quality. > I'm running vm on netapp san though nfs 4.1, and I have very good

Re: [pve-devel] LXC volumes on ZFS over iSCSI

>>NFS does not provide good iops and has the ability to bring down a >>node. IMHO NFS is only useful as filestorage server for backups and >>iso images. Well, this is very dependent of nfs server implementation quality. I'm running vm on netapp san though nfs 4.1, and I have very good

Re: [pve-devel] LXC volumes on ZFS over iSCSI

On Mon, 10 Oct 2016 22:30:04 +0200 (CEST) Alexandre DERUMIER wrote: > > I'm afraid of behaviour if a node is not joinable to do the delete or rescan. > Seem to be difficult to manage with big cluster. > I'm curious to see result of: > It seems to work :-) 1) After login

Re: [pve-devel] LXC volumes on ZFS over iSCSI

On Tue, 11 Oct 2016 21:26:08 +0200 (CEST) Dietmar Maurer wrote: > > Why not NFS? This would make everything easier. IMHO iSCSI is really clumsy. > NFS does not provide good iops and has the ability to bring down a node. IMHO NFS is only useful as filestorage server for

Re: [pve-devel] LXC volumes on ZFS over iSCSI

> > VMs use KVM live backup feature, which is not available for containers. > > > I see. A work around could be to make a clone of a snapshot which is > then exposed through iscsi. Would that be an idea? Why not NFS? This would make everything easier. IMHO iSCSI is really clumsy.

Re: [pve-devel] LXC volumes on ZFS over iSCSI

On Tue, 11 Oct 2016 19:54:05 +0200 (CEST) Dietmar Maurer wrote: > > VMs use KVM live backup feature, which is not available for containers. > I see. A work around could be to make a clone of a snapshot which is then exposed through iscsi. Would that be an idea? --

Re: [pve-devel] LXC volumes on ZFS over iSCSI

> > Besides, iSCSI has many other drawbacks, for example it is not possible > > to access ZFS snapshots over iSCSI. If we use ZFS/NFS instead, we can have > > all that functionality? > > > To what purpose is it needed to be able to access a ZFS snapshot? We need that for vzdump container

Re: [pve-devel] LXC volumes on ZFS over iSCSI

On Tue, 11 Oct 2016 17:39:02 +0200 (CEST) Dietmar Maurer wrote: > > Besides, iSCSI has many other drawbacks, for example it is not possible > to access ZFS snapshots over iSCSI. If we use ZFS/NFS instead, we can have > all that functionality? > To what purpose is it needed

Re: [pve-devel] LXC volumes on ZFS over iSCSI

> I think this is highly hypothetical since a LUN at any point in time > can only be active on one node, proxmox that is, so the hole operation > is serializable which means that every step will be controlled and can > be rolled back. Eg. we are dealing with a deterministic state machine.

Re: [pve-devel] LXC volumes on ZFS over iSCSI

On Tue, 11 Oct 2016 08:33:02 +0200 (CEST) Alexandre DERUMIER wrote: > > I agree with Dietmar. > > Think of a missed ssh command to remote node (small network timeout for > example), for a resize volume for example. > > the lun will still be there, but wrong size, and if

[pve-devel] RFC: live local storage migration

Hi, This is a first try to implement live local storage migration at the same time than vm migration. some users have made requests about it https://forum.proxmox.com/threads/when-will-kvm-live-suspend-migration-on-zfs-work.29049/ This patches serie implement it with 1 disk local disk

[pve-devel] [PATCH 3/3] drive_mirror : add support for nbd uri target

Signed-off-by: Alexandre Derumier --- PVE/QemuServer.pm | 21 +++-- 1 file changed, 15 insertions(+), 6 deletions(-) diff --git a/PVE/QemuServer.pm b/PVE/QemuServer.pm index b80f8f1..bbee7cd 100644 --- a/PVE/QemuServer.pm +++ b/PVE/QemuServer.pm @@ -5800,16

[pve-devel] [PATCH 2/3] add targetstorage to vm_start

This will create a new drive for first local drive found, and start the vm with this new drive. a nbd server is started in qemu and expose volume to network port Signed-off-by: Alexandre Derumier --- PVE/API2/Qemu.pm | 14 ++- PVE/QemuServer.pm | 75

[pve-devel] [PATCH 1/3] add live storage migration with vm migration

This allow to migrate a local storage (only 1 for now) to a remote node storage. When the target node start, a new volume is created and exposed through qemu embedded nbd server. qemu drive-mirror is done on source vm with nbd server as target. when drive-mirror is done, the source vm is

[pve-devel] [PATCH common] SectionConfig: always write out booleans

Because when only writing the ones which are true we're breaking the ones which default to true (like the mkdir option on directory storages, where we need a false value to be written out explicitly). --- src/PVE/SectionConfig.pm | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git

Re: [pve-devel] [RFC container 1/2] fix #1147: allow marking non-volume mps as shared

On Tue, Oct 11, 2016 at 08:24:24AM +0200, Fabian Grünbichler wrote: > On Tue, Oct 11, 2016 at 07:11:45AM +0200, Dietmar Maurer wrote: > > > On October 10, 2016 at 1:04 PM Fabian Grünbichler > > > > > > wrote: > > > > > > > > > this introduces two new options for

Re: [pve-devel] LXC volumes on ZFS over iSCSI

> posses another problem which will be a problem for other things too and > must be handled by fencing. not really that easy - we have a quorum system to handle most situations ... fencing is only required for some special cases. ___ pve-devel mailing

Re: [pve-devel] LXC volumes on ZFS over iSCSI

>>Offline nodes is not a problem because when the get online their scsi >>bus will not have references to disappeared luns. Not reachable nodes >>posses another problem which will be a problem for other things too and >>must be handled by fencing. I agree with Dietmar. Think of a missed ssh

Re: [pve-devel] [RFC container 1/2] fix #1147: allow marking non-volume mps as shared

On Tue, Oct 11, 2016 at 07:11:45AM +0200, Dietmar Maurer wrote: > > On October 10, 2016 at 1:04 PM Fabian Grünbichler > > > > wrote: > > > > > > this introduces two new options for non-volume mount points, > > modeled after the way we define 'shared' storages: > > -

Re: [pve-devel] LXC volumes on ZFS over iSCSI

On Tue, 11 Oct 2016 06:14:14 +0200 (CEST) Dietmar Maurer wrote: > > Such things cannot work, because nodes can be offline (or worse, online > but not reachable). > Offline nodes is not a problem because when the get online their scsi bus will not have references to