Using xCAT 2.8.4 with RHEL 6.5
Trying to get xCAT to setup the Infiniband card and having trouble.
I set up my node definitions as I did in xCAT 2.7.5 (on a different cluster),
namely like this:
lsdef node1:
.
installnic=eth0
otherinterfaces=node1-c:198.18.12.254,node1-ib:198.18.8.254
Where the management interfaces are named <node>-c and the Infiniband
interfaces are names <node>-ib.
What's not happening is the setup of "ifcfg-ib0" on the machine once it's
imaged.
I found the 'configib' script but it relies on a bunch of environment variables
to do its thing.
A little more digging and I found this page:
http://sourceforge.net/p/xcat/wiki/Configuring_Secondary_Adapters/
Which discusses the new 'nics' table and the role of the various columns.
Mine was initially empty, but trying to follow along with the page above, I
added the following:
#node,nicips,nichostnamesuffixes,nichostnameprefixes,nictypes,niccustomscripts,nicnetworks,nicaliases,comments,disable
"node1","ib0!198.18.8.254","ib0!-ib",,"Infiniband",,,,,
Apparently one needs to run the 'confignics' postscript which will run the
'configib' script.
Still, under this setup, the 'configib' script does not appear to get all the
values it needs, specifically the IP address that the
Infiniband card needs.
Note this output that I added to the script:
node1: nic=ib0, nicip=
node1: nic=ib0, nicnet=
I'll keep digging but any help with getting Infiniband setup would be
appreciated.
[root@head postscripts]# updatenode node1 -P 'confignics --ibaports=1'
node1: export: Command not found.
node1: export: Command not found.
node1: Mon Nov 10 12:25:24 EST 2014 Running postscript: confignics --ibaports=1
node1: confignics on node1: config install nic:0, remove: 0, iba ports: 1
node1: NICIPS=ib0!198.18.8.254
node1: confignics on ln116254: executed script: configib for nics: ib0, ports: 1
node1: Low level hardware support loaded:
node1: mlx4_ib
node1:
node1: Upper layer protocol modules:
node1: ib_ipoib
node1:
node1: User space access modules:
node1: rdma_ucm ib_ucm ib_uverbs ib_umad
node1:
node1: Connection management modules:
node1: rdma_cm ib_cm iw_cm
node1: Configured IPoIB interfaces: none
node1: Currently active IPoIB interfaces: none
node1: Unloading OpenIB kernel modules:[ OK ]
node1: Loading OpenIB kernel modules:[ OK ]
node1: error: "net.bridge.bridge-nf-call-ip6tables" is an unknown key
node1: error: "net.bridge.bridge-nf-call-iptables" is an unknown key
node1: error: "net.bridge.bridge-nf-call-arptables" is an unknown key
node1: net.ipv4.ip_forward = 0
node1: net.ipv4.conf.default.rp_filter = 1
node1: net.ipv4.conf.default.accept_source_route = 0
node1: kernel.sysrq = 0
node1: kernel.core_uses_pid = 1
node1: net.ipv4.tcp_syncookies = 1
node1: kernel.msgmnb = 65536
node1: kernel.msgmax = 65536
node1: kernel.shmmax = 68719476736
node1: kernel.shmall = 4294967296
node1: nic=ib0, nicip=
node1: nic=ib0, nicnet=
node1: No network defined for ib0
node1: Unloading OpenIB kernel modules:[ OK ]
node1: Loading OpenIB kernel modules:[ OK ]
node1: Postscript: confignics --ibaports=1 exited with code 0
node1: Running of postscripts has completed.
node1: export: Command not found.
node1: DSH_TARGET_RC: Undefined variable.
Error: node1 remote shell had exit code 1.
------------------------------------------------------------------------------
_______________________________________________
xCAT-user mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/xcat-user
