Nginx VM
-
@Pete-S said in Nginx VM:
I think 20GB is outrageously large. 8GB or preferably 4GB.
Kind of, but thin provisioning means that we could also say that 8GB is unreasonably small. Why not have the buffer room "just in case" since it costs nothing?
-
@Dashrender said in Nginx VM:
@Pete-S said in Nginx VM:
I think 20GB is outrageously large. 8GB or preferably 4GB.
Thin provision though - so what does it really matter?
That's why we do it, avoid unnecessary limitations just in case a log explodes or we need to move files around or decide to change how something is used in the future. I've seen too many installs over the years have issues because they cut it as close as projected but then needed just a little extra space.
-
I saw that a lot on Ubuntu installs - The /boot partition that the installer created was always too small for the kernels after a few updates. Have yet to see that happen in Fedora.
-
@brandon220 said in Nginx VM:
I saw that a lot on Ubuntu installs - The /boot partition that the installer created was always too small for the kernels after a few updates. Have yet to see that happen in Fedora.
Fedora handles that space completely different. No issues there. Current Ubuntu does not, either. I prefer Fedora, but Ubuntu is fine. Issues of old versions don't affect the current one.
-
I had a heck of a battle last night with KVM. I converted to VMs built in HV 2019 to KVM on Fedora 29. Had to use dracut and reload the kernel to get them to boot. Took a while as I am not an expert by any means with KVM. I am upgrading a server tonight and still am on the fence about which hypervisor to go with on the new build. I like KVM because it is "lightweight" and has not given me any grief on guests that were built on the host and not migrated to it. I have a handful of HV hosts and no issues either. This new install will only be running Fedora guests. Leaning more towards KVM. I wish there was an "export" feature built into virt-manager.
-
@scottalanmiller said in Nginx VM:
@brandon220 said in Nginx VM:
I saw that a lot on Ubuntu installs - The /boot partition that the installer created was always too small for the kernels after a few updates. Have yet to see that happen in Fedora.
Fedora handles that space completely different. No issues there. Current Ubuntu does not, either. I prefer Fedora, but Ubuntu is fine. Issues of old versions don't affect the current one.
But I also have no reason to switch back to using it. Those issues are what drove my preference for CentOS back in the day.
-
@brandon220 said in Nginx VM:
This new install will only be running Fedora guests. Leaning more towards KVM. I wish there was an "export" feature built into virt-manager.
KVM really shines for Windows, actually. For pure Linux guests, the playing field is more level, rather than somehow weighted towards KVM. But if you have any Windows, then it is weighted more strongly towards KVM. KVM is the performance leader for Windows workloads.
But from your description, you have benefits to KVM, and not to Hyper-V. Your description of Hyper-V is only that it is "not that bad" or "good enough". But that's not the same as a benefit. Unless you have clear, solid benefits on Hyper-V to offset the benefits of KVM, is there really even a choice to make? KVM seems the absolute unquestioned winner based on your description. Not a winner by much, but a clear winner nonetheless.
-
@brandon220 said in Nginx VM:
I had a heck of a battle last night with KVM. I converted to VMs built in HV 2019 to KVM on Fedora 29. Had to use dracut and reload the kernel to get them to boot. Took a while as I am not an expert by any means with KVM. I am upgrading a server tonight and still am on the fence about which hypervisor to go with on the new build. I like KVM because it is "lightweight" and has not given me any grief on guests that were built on the host and not migrated to it. I have a handful of HV hosts and no issues either. This new install will only be running Fedora guests. Leaning more towards KVM. I wish there was an "export" feature built into virt-manager.
When you migrate, you always have to rebuild that.
https://mangolassi.it/topic/8351/updating-the-linux-boot-image-before-migrating-from-vmware-to-hyper-v -
@JaredBusch I did have to research it. None of the guides I followed for conversion mentioned it though. I was skeptical that it would work without intervention. It was definitely a learning exercise.
-
@scottalanmiller I'll say it again - I wish KVM had an export function that was a "one-step" process. I know I can do a clone in virt-manager but I want to be able to move the image to a new host, or for just a crude backup. Seems like every other hypervisor has that built in.
What is the proper method to move to another host without shared storage? I know you can create the xml and tar a copy but there has to be an "easier" way.
-
@brandon220 said in Nginx VM:
@scottalanmiller I'll say it again - I wish KVM had an export function that was a "one-step" process.
I think KVM lacks "one step" because it's so simple that no one bothers? You could write a script in 30 minutes to make it one step if you wanted.
-
@brandon220 said in Nginx VM:
I know you can create the xml and tar a copy but there has to be an "easier" way.
That's so easy, I'm not sure what an "easier way" would really look like.
-
@scottalanmiller You are right. It is simple. I just worry that I will mistype a command and corrupt the guest. I worry a lot and over-think things way to often.
-
@brandon220 said in Nginx VM:
@scottalanmiller You are right. It is simple. I just worry that I will mistype a command and corrupt the guest. I worry a lot and over-think things way to often.
I see. Yeah, it's not a GUI.
Does oVirt not offer that? I don't use that to know.
-
@scottalanmiller Don't know. Never looked into it.
-
Another question - Is there any reason NOT to use UEFI boot on all guests in KVM?
-
@brandon220 said in Nginx VM:
@scottalanmiller Don't know. Never looked into it.
What system are you using to "make KVM easy"? oVirt is probably the easiest overall and is where most of the "easy features" come from. Cockpit is by far the easiest for the extreme basics, but lacks enough to be problematic for most.
If you are doing everything manually from the command line, that's why it would seem harder. Compare to Hyper-V from PowerShell only. Both are fine that way, but I bet you'll find KVM way easier.
Then if you are using Hyper-V GUI tools, compare to oVirt as KVM's easy cluster management tool. I bet KVM is again, easier.
-
@scottalanmiller I've just been using virt-manager and some ssh. I have Cockpit running on the host but haven't used it for management. It doesn't seem to reflect the guests that are running in the Cockpit gui. The cockpit-machines package is installed.
-
@brandon220 said in Nginx VM:
It doesn't seem to reflect the guests that are running in the Cockpit gui. The cockpit-machines package is installed.
No, you have to either configure them for that, or build them through Cockpit for them to be seen there.
-
@brandon220 said in Nginx VM:
@scottalanmiller You are right. It is simple. I just worry that I will mistype a command and corrupt the guest. I worry a lot and over-think things way to often.
Same thing could happen if you mistype a PowerShell command on Hyper-V.
