Thomas, Sorry, I do not quite follow your point about "2)", is the observed behavior different from what you expected ? As to "duplicate lease" point, is it possible you have an IP associated with a MAC, that is also inside dynamic range ? Can you show: lsdef -t network -l lsdef <node> -i ip,mac makedhcp -q node Mark Gurevich Poughkeepsie Development Lab HPC Software Development - xCAT "If we knew what it was we were doing, it would not be called research, would it?" --Albert Einstein From: "Thomas HUMMEL" <thomas.hum...@pasteur.fr> To: <xcat-user@lists.sourceforge.net> Date: 09/30/2021 02:04 PM Subject: [EXTERNAL] Re: [xcat-user] xCAT 2.16.2 new xNBA issue looking at the dhcp.pm plugin code I wonder if this behavior is not expected, i.e. boot file name is empty in the event of a boot (as in the chain table and as opposed to an install) : is it so ? Thanks for your help -- Thomas HUMMEL On 30/09/2021 17:39, Thomas HUMMEL wrote: > > > On 23/09/2021 15:12, Mark Gurevich wrote: >> Thomas, have you tried using iPXE version 2.21.1 as suggested by >> kcgthb in >> https://github.com/xcat2/xcat-core/issues/7036 $ > > > Hello, > > I tried the rpm you provided in the above issue : > > # md5sum xnba-undi-1.21.1-1.noarch.rpm > 88cfc461e92d67a0a7347345d8c64adb xnba-undi-1.21.1-1.noarch.rpm > > it indeed works for: > > physical netboot > physical discover (genesis) > > but I still have the following strange behavior with a stateful VMWare > VM (UEFI boot): > > > starting a freshly created VM I: > > 1) > > - configure boot order : network first, disk second > - nodeset <node> osimage=<osimage> > - boot the VM > -> I can see tftp of xNBA in MN logs > -> tcpdump shows the expected boot file name (xnba.efi) in the DHCPOFFER > response > - installation starts > - machine ends up, as expected with boot order inverted (disk first, > network second) > > 2) I then just change again boot order in UEFI firmware (network first, > disk second) > > - reboot the machine > -> console shows Attempting to start up from EFI network... > -> no tftp in MN logs > -> tcp dump shows 'boot file name not given' in the several DHCPOFFER > responses > - firwmare eventually ends up booting on the second target (disk) after > some timeout I guess > > After 1) as I always experienced, only the file without the .uefi > extension gets changed *to run only 'exit' in order for the stateful > machine to just pass on to the next target I guess) in > /tftpboot/xcat/xnba/nodes/<node> : the <node>.uefi file still contains > the iPXE script instructing to fetch osimage...which seesm dangerous to > me for stateful install in the event the latter would get http fetched, > but I guess this is another topic. > > Note: > > logs complain about duplicate lease > > 2021-09-30T17:28:06.893678+02:00 maestro-xcat dhcpd[1966303]: uid lease > 192.168.145.126 for client 00:50:56 > :b5:ac:de is duplicate on eth0 > > but from what I see in the leases file, 2 entries match the mac: the > 'static' one (matching the xCAT node's mac attribute and associated to > the node expected ip address, i.e. coming from makedhcp <node>, and one > in the pool of dynamic range ip addresses but with an expired end date. > > Can you help me figuring out what's going on ? > And thanks again for the new xNBA which fixes most of the previous issues. > > -- > 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
_______________________________________________ xCAT-user mailing list xCAT-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xcat-user
