On 2/12/19 4:16 PM, Er Tao Zhao wrote:
Hi，Thomas
Hello, thanks for your answer.
Really long mail. ^_^
Sorry ;-)
Your scenario is normal in a HPC cluster.
eth0 ==> carry both BMC and host data means it running in SHARED mode.
Yes, that's what we did when we had only 1Gb/s connections.
So, I'd like to make sure that eth0 and eth2 are in the same subnet.
Yes. Same vlan, same subnet. We just arranged that BMC tagged packets
and the port was correctly set up in trunk mode. Data packets are
untagged by the OS.
Then, my explanation are blue as below.
But, I strongly suggest you to update to latest xCAT 2.14.5+patch
https://github.com/xcat2/xcat-core/issues/5905#issuecomment-448165125 which
have enhanced a lot for hardware discovery progress.
Yes, I'm using the last one for the next cluster, which I upgrade with
go-xcat (very convenient by the way!)
So, basically, I don't understand the conceptual difference between 1)
and 4) and thus why in 1) nodeA gets discovered without the need to run
nodediscoverstart (granted it has switch and swithport attributes but
not refering the 1Gb/s switch)...
Er Tao: for 1, since there is node: nodeA defined in xCAT database, and
the mtm and serial have been updated by switch based node discovery,
then discovery request can be matched with MTMS based discovery.
Oh I get it : so it's like a fallback from switch-based to mtms like
scenario, correct ?
No
rmdef nodeA
What do you mean here ?
for 4, when a node need to be matched with sequentially, xCAT
will find free nodes specified by nodediscoverstart noderange=foobar
Ok, so basically, nodediscoverstart goal is to specify the free node
pool, right ? But it doesn't seem to work if started too late : if I did
not start it before booting the CN, it won't work, event if the node PXE
looped. That I didn't understand.
Er Tao: If you update xCAT newer that 2.12.4, there is an option
`noboot` for site.dhcpinterfaces to specify the NIC on MN that not reply
DHCP configuration file for the DHCP request get from this NIC. So, it
means, if MN eth0 connect to CN eth0/BMC, eth1 --> CN eth2.
Then, with the configuration below and `mknb <ppc64|x86_64>`, the CN
won't get DHCP configuration from eth0.
chdef -t site dhcpinterfaces="eth0:noboot,eth1"
I'm not sure I get this : eth0 and eth1 in this statement are MN nics
right ?
Since CN eth0 and eht2 are on the same subnet in my case, that would not
work, or did I misunderstood something ?
Thanks!
--
Thomas H.
_______________________________________________
xCAT-user mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/xcat-user
