I have been using proxmox for a few years now and I am wondering why there is still no supported HA solution for containers? At this point I an willing to code it myself....(i am a sysadmin though). This feature would literally phenomenal.... Is there some sort of theoretical limitation?
why no Container HA
- Thread starter thesubmitter
- Start date
HA for container is available since years. seems you missed that.
http://pve.proxmox.com/wiki/High_Availability_Cluster
http://pve.proxmox.com/wiki/High_Availability_Cluster
The current state of support for HA CT is not a true HA solution since the CT will have to be suspended for noticeable time measured in seconds while for KVM the suspended time is measured in milliseconds. True HA for CT will IMHO first be available when we see plop support.
not sure about what you talk here.
our resource manager (rgmanager) checks if the CT or VM is running. If not, rgmanager starts the CT or VM on one of the remaining nodes.
this is not a process of suspending or withing milliseconds. its a simple restart of the CT/VM on a remaining node. of course, the failed node must be successfully fenced. I do not see how ploop can help here.
That wont work unless you have a shareable FS on top of the SAN otherwise you cant mount it in 2 places at once.
I incorrectly always thought the NFS option was a hack and my personal view of NFS is sub-optimal performance ...
I incorrectly always thought the NFS option was a hack and my personal view of NFS is sub-optimal performance ...
Why I cannot setup CLVM+general FS like EXT4 for CTs just as KVM works?
When I setup KVM HA I don't install virtual OS on shared FS like GFS. All work fine with CLVM and usual EXT4 on top.
Clvm = cluster lvm = mountable in 2 locations at once.
I am not sure why only nfs supports HA CTs
I am not sure why only nfs supports HA CTs
Clvm presents a block device to the vm which is fine for kvm but ct needs a file system since it cannot mount block devices on its own. For this to work proxmox must implement plop support for ct.
That is only partly true. With plop you are not limited to NFS but will be able to use any storage supported by proxmox to host HA CT since plop does not require a file system.
From what I have been told the devs are not against supporting plop but waits for the final patches for 3.10 kernel to emerge before making their decision. I find it a bit odd that openvz patches for 3.10 is not available yet - it is more that 6 month since Redhat released RHEL 7 with the 3.10 kernel.
It has already been raised at the forum: http://forum.openvz.org/index.php?t=msg&th=12793&start=0&S=c6c6b1923e3f3d637a997f225e14bae1
But as can be seen no official response yet for over a month! Maybe openvz has seen its days of glory?
Try to translate this article: http://www.opennet.ru/opennews/art.shtml?num=41162
Especially "Проект OpenVZ постоянно усовершенствуется, портируется на новые ядра. Компания добавляет функциональность, улучшает производительность и своевременно выпускает обновления, в том числе связанные с безопасностью. В данный момент мы сосредоточены на стабилизации ядра на базе RHEL7, которое довольно быстрыми темпами приближается к статусу beta."