Djelibeybi
commented
3 years ago I'm busy downloading OpenVZ so I can try and reproduce this issue.
Open centminmod opened 3 years ago
Djelibeybi
commented
3 years ago I'm busy downloading OpenVZ so I can try and reproduce this issue.
Djelibeybi
commented
3 years ago Can you confirm whether you're running a CentOS 8 VM or container on OpenVZ? In fact, if you could share the prlctl commands you used to create the instance, that would be very helpful. I've never used OpenVZ before, so any info you can provide to overcome the learning curve would be appreciated!
centminmod
commented
3 years ago I am not the one managing the OpenVZ VPS host node. I just rented a OpenVZ VPS from https://hostus.us/openvz-vps.html with CentOS 8 image out of the box and then ran yum update to get to CentOS 8.3 and then ran cento2ol.sh script. OpenVZ VPS use the Kernel from the host node unlike KVM/non-OpenVZ virtualized VPS servers https://hostus.us/openvz-vs-kvm.html.
OpenVZ uses a common shared kernel. Because of the shared kernel, overhead is lower and so is RAM consumption per VPS. A disadvantage to OpenVZ would be that a kernel failure could cause a failure across all VPS instances running on the same server. The shared kernel limits the operating systems you can run and keep in mind, the latest OSes with newer kernels may not work under OpenVZ.
Just quickest way to test is just rent a OpenVZ VPS and choose CentOS 8 as the out of the box image from say https://hostus.us/openvz-vs-kvm.html or another OpenVZ provider.
For scripting in centos2ol.sh you could use output from virt-what to determine if the server environment is openvz/lxc based in which case the host node's Kernel is used
On Hostus.US's OpenVZ VPS, output is
virt-what
openvz
lxcOn KVM server it would output = kvm and dedicated would have empty output
Djelibeybi
commented
3 years ago Gotcha. The script (currently) assumes its running on an instance with a dedicated kernel. We hadn't considered switching a container because it would generally be easier just to recreate one. But that ignores the VPS use case.
Can you confirm that your instance is switched and still works? As there is no kernel, the dracut error should have been purely cosmetic.
centminmod
commented
3 years ago Oh hadn't checked as I reinstalled CentOS 8 again as I assumed the conversion wasn't complete or was it complete and just dracut error can be ignored ?
Djelibeybi
commented
3 years ago Hopefully the latter, given dracut isn't needed in this scenario and it's the last thing the script does.
Djelibeybi
commented
3 years ago It's not a fix, but we've added specific docs around the need to be using a CentOS kernel (for now).
centminmod
commented
3 years ago Thanks @Djelibeybi for the update. You could add to centos2ol.sh a check for output from virt-what to see if lxc/openvz is detected to abort the script with a notice before it goes too far in the process?
Djelibeybi
commented
3 years ago Yes, that's the longer-term plan, which requires access to a testing environment and unfortunately we can't just rent third-party OpenVZ based instances, because it would all have to go through procurement, etc. I'm still working on getting a local install of OpenVZ running so I can test on that.
On OpenVZ 7 VPS, the centos2ol.sh script will fail with below output for
dracut: Cannot find module directory /lib/modules/4.18.0/