On Tue, Sep 8, 2020 at 9:50 AM <harryo.dk@gmail.com> wrote:
Hi, why does my ovirt hypervisor node crash when i pull the hypervisor boot drive? The following thread claims that the oVirt node hypervisor should lodad into memory, so boot disk failure shouldn't have any impact on a running node. https://lists.ovirt.org/pipermail/users/2014-December/029771.html
Because the above link is old, and refers to what is now called "oVirt node Legacy" (or in some places "vintage"). In 4.0 we moved to what's now called "oVirt node Next Generation", or "NG" for short. Search the net for e.g. "ovirt node ng" "slides" to find some presentations about this change.
Only memory failure should have a impact, and if RAM fails there is a possibility that the hypervisor OS would have crashed regardless. If you load the hypervisor OS into memory there will be moved/combined a point of failure (the boot disk) with another(the RAM).
You are welcome to file an RFE in bugzilla for this. AFAIK this is not supported in ovirt-node currently. If you want a diskless ovirt host/hypervisor, you are more than welcome to try any of the existing methods you can find on the net for installing and using CentOS diskless, and then use that as an oVirt host. Relevant methods might include SAN boot (if your hardware supports that), NFS root, etc. Not sure about root on tmpfs, that might be possible as well. Best regards, -- Didi