New KVM 1.2, snapshots, vzctl and other packages on pvetest (on the road to V2.2)

[martin]

We just finished uploading a lot of new packages to the pvetest repository, including latest KVM 1.2 and live snapshots. OpenVZ toolset is updated, introducing a new console view. Especially for beginners it is not that easy to understand and manage containers but with the new console this is big step forward (OpenVZ and KVM console looks now quite similar).

Everybody is encouraged to test and give feedback as these packages will be the base for V2.2 release.

Release notes
- pve-manager (2.2-21)

• openvz: use real console instead of 'vzctl enter'
• openvz: remove init-logger pagers
• add workaroud to correctly show scrollbars
• add snapshot support to GUI
• vzrestore: allow --storage option
• fix bug 178: correctly verify if VMID is already used
• GUI: allow to select SCSI disks when creating VMs (booting from SCSI disks works again)
• Slovenian translation (thanks to Dejan Borstnar)
• new option to select SCSI controller hardware
• add SATA to GUI

- ibpve-storage-perl (2.0-33)

• create 'dump' directory if storage contains containers
• qcow2 : preallocate metadatas at file creation
• nexenta: implement storage status, several fixes
• add snapshot support

- pve-kernel-2.6.32 (2.6.32-78)

• update to vzkernel-2.6.32-042stab061.9.src.rpm
• update e1000e to 2.1.4
• update ixgbe to 3.10.17
• remove ixgbe_fix_mac_flush.patch (now upstream)

- qemu-server (2.0-59)

• fix bug 251: use new command line syntax
• include snapshot support
• disable vzdump for VM containing snapshots
• support up to 32 network devices
• support up to 16 virtio devices
• scsihw: add megasas controller
• implement 'qm rescan' to update disk sizes and unused disk info

- pve-qemu-kvm (1.2-6)

• update to 1.2.0
• add patch for snapshot support

- vzctl (4.0-1pve2)

• depend on attr package (scripts use setfattr)
• update to latest vzctl
• removed init logger patches

- other updates

• update vzquota to 3.1
• update redhat-cluster-pve to 3.1.93
• update fence-agents to 3.1.9
• update corosync to 1.4.4
• update libiscsi to 1.6.0
• update ceph packages to 0.48.2argonaut
• update sheepdog to 0.5.0

About Proxmox VE package repositories
http://pve.proxmox.com/wiki/Package_repositories

 

[Marek Panek]

live snapshots

Is it livebackup (incremental) ?

 

[tom]

live snapshot is no backup.

you can save the current VM state (KVM only), including RAM. and you can rollback later

Description will follow, see http://pve.proxmox.com/wiki/Live_Snapshots

 

[jforeman]

live snapshot is no backup.

you can save the current VM state (KVM only), including RAM. and you can rollback later

Description will follow, see http://pve.proxmox.com/wiki/Live_Snapshots

So in that case what will be used for qemu backups? You have this listed under the changes for qemu-server:

• disable vzdump for VM containing snapshots

 

[dietmar]

So in that case what will be used for qemu backups?

Nothing. vzdump backups simply does not work when you use VM snapshots (for now).

 

[badji]

Bonsoir,
Thank you very much.
I'm updating my three servers. no bug for now.
I have not yet tested sheepdog

 

[Jae Lee]

i'm thinking of upgrade as i can really take great use of the live snapshot feature.
so if i'm reading correct, proxmox now supports the XenServer/VMWare-like snapshots where it takes only a few seconds for VMs of any size all while using up considerably less disk space?

also, how feasible is it to perform this upgrade on a live server? i've never done proxmox upgrades before so i'm a bit worried what can go wrong (if any) during the upgrade process.

 

[lozair]

Reading the source code it seems snapshot was possible only for specific storage (sheepdog etc..) or disk format like qcow2.
Do you plan to implement snapshot for vm disk hosted on LVM volume ?

Regards

 

[dietmar]

Do you plan to implement snapshot for vm disk hosted on LVM volume ?

There are plans, but nothing we do in next time. Do you talk about local LVM or LVM on shared storage?

 

[mir]

OpenVZ toolset is updated, introducing a new console view.

Does this means that online migration is now able to succeed even when a console to a VZ is open?

 

[dietmar]

Does this means that online migration is now able to succeed even when a console to a VZ is open?

yes

 

[lozair]

I was talking about shared storage

 

[RCK]

In proxmox 2.2, would it be possible to update the kernel module DRBD to 8.3.13 ?
https://bugzilla.proxmox.com/show_bug.cgi?id=185

Thanks in advance,

 

[snowman66]

I was testing sheepdog cluster lately on 3 nodes pve cluster. If i disconnect one of the nodes the sheepdog storage is not usable any more and data can be lost. I guess this is expected in current version?

 

[tom]

sheepdog is not stable and not recommended now, and yes, such issues are expected. (Ceph RBD is considered to be more stable, but also nothing for production use for now)

 

[dietmar]

In proxmox 2.2, would it be possible to update the kernel module DRBD to 8.3.13 ?
https://bugzilla.proxmox.com/show_bug.cgi?id=185

The OpenVZ team is working on that: http://bugzilla.openvz.org/show_bug.cgi?id=2254

 

[MaFL]

Can I run stable and testing in a cluster?

Regards
Matt

 

[dietmar]

Can I run stable and testing in a cluster?

No, that is not tested.

 

[TomTomGo]

There are plans, but nothing we do in next time. Do you talk about local LVM or LVM on shared storage?

Will it be based on lvcreate / lvconvert commands ?