Posts made by marekm

According to wikipedia: "AMD Opteron CPUs beginning with the Family 0x10 Barcelona line, and Phenom II CPUs, support a second generation hardware virtualization technology called Rapid Virtualization Indexing (formerly known as Nested Page Tables during its development), later adopted by Intel as Extended Page Tables (EPT). " - so that would be since about 2008 or so.

Later there have been some critical microcode updates, 63xx series got one in 2016 (unprivileged guest could crash the host), not sure which older CPUs have this update and which don't.

What about future support for old AMD CPUs? I see Opteron 61xx and older support was removed in CH 8.0, are there any known issues with 62xx/63xx/P series that might cause them to become unsupported in the near future?
I rely on used servers mainly for cost reasons, recently moved from Intel to AMD in the hope they will be less insecure (not perfect either, but what is), current 6338P CPUs from 2014 seem to be good enough for now, until shiny new EPYC become cheaper in a few years...

@stormi Thanks. Is this https://github.com/xenserver/gpumon - if yes, it doesn't look like something big, perhaps could be patched to remove dependency on proprietary nvidia stuff.

@stormi OK, just asked since the issue is about a year old and still unanswered by Citrix, wondering why it didn't show up in earlier xcp-ng releases. Perhaps they have only recently made some change to XAPI so it requires gpumon, and this change could be reverted (or better yet, fixed to handle the missing package gracefully, not hang the whole machine at boot time).

@stormi Is this related to https://bugs.xenserver.org/browse/XSO-936 (build dependency on gdk-devel) - or a different issue? (I suspect gdk-devel is some part of GNOME, not nVidia proprietary - but could be wrong.)

Can their improvements be combined with the ones made by XCP-ng for even better performance?

https://docs.citrix.com/en-us/citrix-hypervisor/whats-new.html

I wasn't aware of the hardcoded /usr/local/lib path as everything is running from $HOME/xen-orchestra and node_modules directory full of symlinks was there too. It works now - thanks!

Testing xcp-ng 8.0 on two servers with ha-lizard-noSAN, so far so good. Now, where to find more settings of openvswitch? For the redundant replication link (DRBD/iSCSI) between two machines without a switch, I'd like to set up a bridge with (R)STP between one 1Gb and one 10Gb interface, where 1Gb is a backup if 10Gb link fails. The ha-lizard reference design calls for bonding two 1Gb interfaces with LACP and this was easy to set up in XenCenter, but bonding requires interfaces of the same speed.

I have tried to follow the instructions, twice (on 64-bit Debian 9.9, and again after upgrade to Debian 10, as VM under XCP-ng 8.0) with the same results each time - basically XO works, but:

• Settings / Plugins - "No plugins found" (I'd like to look at xo-server-sdn-controller for example)
• Settings / Config - "No config import available for Community Edition"
  Are there some important steps I missed, or is some $$$ required to make that work? Thanks!