Very slow imaging performance on XCP-NG guest vm

[DBCountMan]

Not a critical issue as I use XCP-NG to image vms for testing the images. Was just wondering if anyone else has this issue.

Whenever I deploy an image to a vm, it crawls down to less than 1GB/min with several delays between screen updates. Eventually I see these message appear:

The host is a Dell PowerEdge R710. I give the guest 4 cores and 8GB ram. I tried 1 socket, 4 cores and also tried 4 sockets 1 core. The network connection is 1Gbps. I tried imaging the same vm using Acronis True Image which took about 8min. If I deploy an image with FOG on the same vm, the process can take +20min.

[Sebastian Roth]

@brakcounty Which version of FOG do you use and even more important, which FOS kernel version is installed?

[DBCountMan]

@Sebastian-Roth FOG version 1.5.10. Not sure how to find the FOS kernel version. Are you referring to the bzImage kernel version? If so the kernel version I’m on is 5.15.93.

[Sebastian Roth]

@brakcounty said in Very slow imaging performance on XCP-NG guest vm:

Are you referring to the bzImage kernel version? If so the kernel version I’m on is 5.15.93.

Yes. Just a hint: You can find out the version of kernel binaries by running file /var/www/html/fog/service/ipxe/bzImage* (or using whatever kernel filename you want to know the version).

We are planing to bring out a fresh new 6.1.x kernel for testing in the next week. Keep an eye on https://github.com/FOGProject/fos/releases for it to come out.

No idea if that newer kernel version will solve the described issue but it’s worth a try.

[Sebastian Roth]

@brakcounty A 6.1.22 test kernel is available on Github. Be aware those kernels are untested so far. I dod not have the time to even do the simplest bootup test.

[DBCountMan]

@Sebastian-Roth The devel kernel version 6.1.22 didn’t help, BUT, I was using the server’s local storage to store the virtual disk. I mounted an SMB share to the xen server, stored the virtual disk on that, voila, no more errors.

Conclusion: Nothing to do with FOG.

[Milheiro]

I had that problem with slow imaging. My images were Zstd level 15, and I changed them to Zstd level 5 and the imaging process increased the speed.
Maybe some one can confirm this.

Hope to help.

[DBCountMan]

@Milheiro Zstd level is compression level right? That I’m aware of. Makes sense if you increase compression that it would take longer to capture/deploy. What I was experiencing was linux kernel level errors, only when deploying to a virtual disk that resides on the xenserver’s local storage array. I wonder if it is something like Truenas where it performs worse when you let the storage controller handle the array vs a soft-raid controlled by the OS.