xNBA 1.0.3 was based on a much older version of iPXE, so there are likely many changes in the iPXE base between xNBA 1.0.3 and 1.20.1. xNBA 1.20.1 is almost identical to iPXE 1.20.1 with minimal customization. The original rationale for updating xNBA to a newer iPXE base was that several hardware platforms were not functional with xNBA 1.0.3 but work better after rebasing xNBA to iPXE 1.20.1. Are there any differences in the BIOS boot mode settings between the servers that are working and the servers that are failing? Legacy mode vs Legacy + UEFI mode vs UEFI only mode? Do you have better results if you downgrade xNBA to the previous version? wget https://xcat.org/files/xcat/repos/yum/2.16/xcat-dep/xnba-undi-1.0.3-131028.noarch.rpm rpm -e --nodeps xnba-undi rpm -ivh xnba-undi-1.0.3-131028.noarch.rpm Also, there is a possibly related issue here: https://github.com/xcat2/xcat-core/issues/6518 Any help and feedback that other xCAT x86 users can contribute about which hardware platforms are working better or worse when using xNBA 1.0.3 vs 1.20.1 will help determine the best recipe for the different hardware platforms that are in use and hopefully allow a golden version of xNBA that works well for everyone to be built. From: Thomas HUMMEL <thomas.hum...@pasteur.fr> To: <xcat-user@lists.sourceforge.net> Date: 05/31/2021 12:31 PM Subject: [EXTERNAL] [xcat-user] xCAT 2.16.2 new xNBA issue Hello, I'm using successfully xCAT for stateless CentOS 8.3 nodes provisioning. Last week I upgraded from xCAT 2.16.1 to 2.16.2 without noticing any issue. However today I experienced the following: My nodes are mainly SuperMicro AS-2124BT-HNTR (Big Twin - 2U) systems (H12DST-B motherboard) which PXE boot on an RJ45 nic adaptater which lspci lists as: Intel Corporation Ethernet Controller 10G X550T (rev 01) These nodes, though physicaly identical, has been purchased as 3 batches at 3 point in time. This has always worked fine across several xCAT versions. However, since the 2.16.2 upgrade and because - I reckon - of the new xNBA ("xNBA has been updated to an iPXE 1.20.1 base"): 1) batch 1 PXE boots, downloads xBNA which prints for a very long time (~25s real time) xBNA initializing devices, and when finally finished prints: xNBA 1.0.0+ ... Features... No more network devices -> so the node fails to boot 2) as batch 2 and 3, though identical physically and with the same version of BIOS and BMC firmware (though a different CPLD version), it works (os image ultimately gets fetched) However, the device initialization takes a fairly amount of time too (~10s) and the whole xNBA stage (before osimage fetching) takes very long as well (~28s). Note: before this, a downloading NBP file takes some time too. I tried to search for hints in iPXE repo commit messages but I'm don't know the specifics enough. Do you have any clue about what has been changed to iPXE/xNBA to show such a behavior ? Thanks for your help -- Thomas HUMMEL _______________________________________________ xCAT-user mailing list xCAT-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xcat-user
_______________________________________________ xCAT-user mailing list xCAT-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xcat-user
