AMD EPYC & OmniOSce/Solaris
- Thread starter m4r1k
- Start date
Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
My homelab setup is an esxi 6.5 all in one with an omniosce vm with the onboard 9300-8i passed thru. Been running totally solid for about 6 months or so now. CPU is a 7351P and mobo is a H11SSL-NC - been super happy with it. I'm planning on updating it to 6.7u1 this weekend as a matter of fact if I don't have too many other honey dos. I have never tried bare metal on it other than esxi, but don't see why it wouldn't work with omnios. You might ask on the new omniosce mailing list or illumos list just to be sure.
If you want, I guess I could fire up the bare metal installer and see what I see while I have it down to do the upgrade - you want me to try latest omniosce or LTS? Only thing that sucks for me about EPYC is that the other two machines in my home lab setup are old R720s and you can't vmotion between Ivy Bridge and EPYC so I have to shut everything down to move it over while I upgrade the EPYC machine then shut it down again to move it back.
If you want, I guess I could fire up the bare metal installer and see what I see while I have it down to do the upgrade - you want me to try latest omniosce or LTS? Only thing that sucks for me about EPYC is that the other two machines in my home lab setup are old R720s and you can't vmotion between Ivy Bridge and EPYC so I have to shut everything down to move it over while I upgrade the EPYC machine then shut it down again to move it back.
Hey BullCreek,
If you could give it a try it would be super sweet!
My current system is based on Intel Xeon v4 and I just want to know if AMD could be an option for the future
Thanks!
Finally got around to upgrading to esxi6.7 today - but before I did that, I tried R151028 on it. Seems to work fine other than a few things I noticed:
1. The installer will boot off USB fine, but if you install to a USB stick, it won't boot after you are done. I think it is this bug:
Bug #8679: failed to attach AMD IOMMU - illumos gate - illumos
2. I wasn't smart enough to figure out how to make it boot off the onboard m.2 drive - again, it would install fine, but wouldn't boot afterward. Someone with more UEFI shell fu than me could probably make it work though.
Finally, I installed it onto just a normal SATA drive, and it booted fine. I installed nappit and everything seemed to work fine. I could see both the m.2 drive and the optane, as well as import the pool attached to the onboard SAS so I think it would probably work ok, keeping the above in mind.
1. The installer will boot off USB fine, but if you install to a USB stick, it won't boot after you are done. I think it is this bug:
Bug #8679: failed to attach AMD IOMMU - illumos gate - illumos
2. I wasn't smart enough to figure out how to make it boot off the onboard m.2 drive - again, it would install fine, but wouldn't boot afterward. Someone with more UEFI shell fu than me could probably make it work though.
Finally, I installed it onto just a normal SATA drive, and it booted fine. I installed nappit and everything seemed to work fine. I could see both the m.2 drive and the optane, as well as import the pool attached to the onboard SAS so I think it would probably work ok, keeping the above in mind.
Hey!
I wasn’t counting on this anymore.
But if I can draw some conclusions, it seems your (any?) EPYC platform is not mature enough for a ZFS use-case which puts reliability and data-integrity as the first priority. When I see random issues on the I/O, that’s at least the feeling I get.
I guess a few more years needs to be spent making sure the AMD platform works as expected.
Thanks for your detailed feedback!
I wasn’t counting on this anymore.
But if I can draw some conclusions, it seems your (any?) EPYC platform is not mature enough for a ZFS use-case which puts reliability and data-integrity as the first priority. When I see random issues on the I/O, that’s at least the feeling I get.
I guess a few more years needs to be spent making sure the AMD platform works as expected.
Thanks for your detailed feedback!