Kernel upgrade without reboot

[wosp]

Since Linux 4.x kernel there are built-in posibility's to upgrade the Linux kernel without a reboot. Since PVE 4.x uses Linux 4.x kernel, are there plans to support this feature out-of-the-box? And if yes, is there an ETA for this?

I think this feature is less important on HA clusters, but can be very usefull for single nodes.

 

[dietmar]

see http://www.proxmox.com/en/partners/technology-partner/item/kernel-care

 

[wosp]

Thanks. But currently no plans for out-or-the-box integration I guess? I don't like the commercial philosophy of CloudLinux, so I'm not very excited to buy a license from them. Don't get me wrong, it's not that I absolutly don't want to pay for software (or support) I use (I also pay for PVE subscriptions and yearly make donations to different opensource projects), but I don't want to be forced because it's the only option to use the software.

 

[fireon]

When high availability is so important for you, then use DRBD or Cepth. This is an optimal solution. So you are able to updaten or service an other node.
https://pve.proxmox.com/wiki/Ceph_Server
https://pve.proxmox.com/wiki/DRBD9

but yes you also upgrade your kernel online. Never done this with debian. But works out of the box with centos.

 

[wosp]

When high availability is so important for you, then use DRBD or Cepth. This is an optimal solution. So you are able to updaten or service an other node.
https://pve.proxmox.com/wiki/Ceph_Server
https://pve.proxmox.com/wiki/DRBD9

but yes you also upgrade your kernel online. Never done this with debian. But works out of the box with centos.

We only use shared storage on all PVE nodes we have and all our production servers are in a HA cluster. In fact, we even don't have/sell any VM that's not HA.

But, for testing purposes and for others (including some colo customers with own hardware) that run PVE on a single node I think this feature may be very welcome. I'm thinking for the whole PVE project here, not only for whats good for me (however we may save some time with less reboots, even on our HA clusters).

 

[fireon]

We only use shared storage on all PVE nodes we have and all our production servers are in a HA cluster. In fact, we even don't have/sell any VM that's not HA.

Yes this sounds fine . But please do not get me wrong. When you have an customer or an office that build there own server, were is really important and have no time for an reboot, i think there is money enough to build some HA Cluster. When not there is something wrong in there business.

 

[Borut]

see http://www.proxmox.com/en/partners/technology-partner/item/kernel-care

Kernel-care is more kernel-uncare. I installed it almost month ago and is still not working. Support is bad and very slow. They were missing last five PVE kernels. They needed more then a week to fixed this. Now, they are still looking why is my kernel not updated automatically....and this in 30 days try out period...

Looks like they are dropping support for Proxmox...

Do you have any better suggestion?