-
Notifications
You must be signed in to change notification settings - Fork 139
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
List namespaces incorrectly reporting numa node on EL8 #211
Comments
This appears to be a duplicate of intel/ipmctl#156 that was reported in the ipmctl GitHub issues. There's more data in that issue we should compare with this issue to confirm it's the same underlying problem. Namely:
Look at the ProximityDomain within the 'NFIT' sections to see if they're zero (0x0) for all regions. If so, this is a BIOS bug/regression problem. |
Relevant output from
|
Also have an almost identical IceLake system which maps region0 to NUMA-1 and region1 to NUMA-0:
Any suggestions would be welcome on how to resolve. |
Maybe you could at lest try the current ndctl? https://build.opensuse.org/package/binaries/hardware:nvdimm/ndctl/openSUSE_Leap_15.3 |
Maybe you could at lest try the current ndctl? I am on the latest version provided by the distro I am running:
I will try to install a later version by building from source. |
Right, you mentioned Leap but I missed that you mentioned it as the working version. Then it likely works because of some random backports or different build configuration, and building the current version from source should get you somewhere either way. |
I re-provisioned the problem IceLake host with Leap15.3 and upgraded ipmctl to v03 and ndctl to v73 (exact versions below), problem persists (pmem0 on numa1 & pmem1 on numa0):
@sscargal any further ideas? |
The numa_node is no longer zero so the reported problem is solved, right? |
Two systems have been mentioned, one has both pmem mapped to zero (wolf-216) and the other has pmem1 mapped to numa0 and pmem0 mapped to numa1 (wolf-220). Neither issue has been resolved. |
I don't see a problem with pmem1 mapped to numa0 - the device order is not guaranteed. As for both regions mapped to numa0 you did not provide log from the problem system with the current ndctl. |
Thanks for clarifying the device order issue, I haven't had access to the problem system that assigned both regions to numa0, I will try to get access to it and test that today. |
I got access to the system, re-provisioned to leap15.3 and updated ipmctl and ndctl to latest. After removing regions, rebooting, creating regions, rebooting, the same issue persists where both regions are on NUMA 0:
|
Not directly related to intel/ipmctl#156 as ProximityDomain field values are nonzero:
|
The numa node is simply:
|
numa_node reported incorrectly (always zero) in ndctl list -v output (IceLake platform):
Regression is not seen on Leap15 (CascadeLake) host with the same version of ndctl:
Similar to regression seen in CentOS 7 and fixed in CentOS 8: #130
The text was updated successfully, but these errors were encountered: