In sharedtftp=1, it is assuming that /tftpboot is nfs or otherwise shared and updating it in one place suffices. In a dynamic pool scenario this *used* to be unworkable, but now static and dynamic can work so long as xNBA is used on x86 (but no POWER solution for that).
From: Dave Barry <[email protected]> To: xCAT Users Mailing list <[email protected]> Date: 10/13/2011 12:21 PM Subject: Re: [xcat-user] Why are service nodes sent commands for *all* nodes? Thanks! Interestingly enough after setting sharedtftp to 1 and restarting xcatd on the service node (and verifying that /tftpboot is now mounted from MN), a nodeset only returns one response (as if the compute node did not have a service node set at all). Is this expected behavior? I was expecting feedback from both the management node and the service node still. [root@mn ~]# lsdef c1n01 Object name: c1n01 arch=x86_64 bmc=c1n01-bmc bmcport=0 chain=runcmd=standby currchain=boot currstate=netboot centos5.5-x86_64-compute groups=compute,all,ipmi initrd=xcat/netboot/centos5.5/x86_64/compute/initrd-stateless.gz installnic=eth0 interface=eth0 ip=192.168.1.2 kcmdline=imgurl=http://!myipfn!/install/netboot/centos5.5/x86_64/compute/rootimg.gz XCAT=!myipfn!:3001 ifname=eth0:00:50:56:11:11:15 netdev=eth0 kernel=xcat/netboot/centos5.5/x86_64/compute/kernel mac=00:50:56:11:11:15 mgt=ipmi netboot=pxe nfsserver=mn nodetype=osi ondiscover=nodediscover os=centos5.5 postbootscripts=otherpkgs postscripts=updaterepos.sh,syslog,remoteshell,syncfiles power=ipmi primarynic=eth0 profile=compute provmethod=netboot servicenode=service01 status=booted statustime=10-11-2011 12:13:27 [root@mn ~]# nodeset c1n01 netboot c1n01: netboot centos5.5-x86_64-compute [root@mn ~]# nodeset c1n01 stat c1n01: netboot centos5.5-x86_64-compute [root@service01 ~]# df -h Filesystem Size Used Avail Use% Mounted on /dev/sda1 33G 1.4G 30G 5% / tmpfs 1006M 0 1006M 0% /dev/shm storage01:/install 47G 6.7G 38G 16% /install storage01:/opt_xcat 47G 6.7G 38G 16% /opt/xcat mn:/tftpboot 57G 7.0G 47G 14% /tftpboot Just verifying there isn't something else in play here I may have missed. On Thu, Oct 13, 2011 at 10:55 AM, Jarrod B Johnson <[email protected]> wrote: If disjointdhcps=0, it largely signals xCAT to not try to make sense of the network in terms of xCAT table relationships beforehand, and asks every xCAT instance to evaluate every part of that request to see if it is applicable to itself. A 'do everything' mode that does more than it has to, but with less configuration required. In disjointdhcps=0, if a node already has an ip address, a static host declaration is set up on the service node if: -the ip address is reachable via a routed netwok, it gets set up on all service nodes -the ip address matches one of the network/netmask combiniations of it's interfaces If a node is dynamic and has no ip address, a host declaration is made everywhere in disjointdhcps=0 case, just in case. Basically, disjointdhcps=1 means to be a bit more surgical in how explicit 'makedhcp' requests are sent out, but when sharedtftp=0, the implicit dhcp refreshes are not so surgical at this moment in time, because the tftp refresh command is currently going out to hit everything because the code does not yet handle all the nuances that the dhcp subsystem does. Inactive hide details for Dave Barry ---10/12/2011 06:27:01 PM---Thanks Linda! I will do this test tomorrow morning once I am bDave Barry ---10/12/2011 06:27:01 PM---Thanks Linda! I will do this test tomorrow morning once I am back in the office and report back. Reg From: Dave Barry <[email protected]> To: xCAT Users Mailing list <[email protected]> Date: 10/12/2011 06:27 PM Subject: Re: [xcat-user] Why are service nodes sent commands for *all* nodes? Thanks Linda! I will do this test tomorrow morning once I am back in the office and report back. Regarding your comment about tftpboot, thinking through the steps, when disjointdhcps is set to 0, the service nodes that are not assigned a particular compute node will never pick up the DHCP request for that node. As a result, they will also never send that node their pxeboot file. I can't think of a reason why the service nodes would then need to ever write out a pxeboot configuration file for a node they will never serve DNS requests to. On Wed, Oct 12, 2011 at 10:24 AM, Linda Mellor <[email protected]> wrote: Dave, I looked into this a little more. In your case you have "sharedtftp=0" set in your site table. So, that means /tftboot is not mounted from the management to your service nodes, so nodeset needs to get sent to all the service nodes to create the correct /tftpboot files. As a quick test, if you set "disjointdhcps=1" and "sharedtftp=1", the nodeset should only go to the service node for that node. What I need to investigate still is why we would need ALL service nodes to have /tftpboot files for ALL nodes, or if this code needs to be made smarter, too. Linda Inactive hide details for Dave Barry ---10/11/2011 02:54:33 PM---Thanks! I'm pretty sure this only applies to service node pool Dave Barry ---10/11/2011 02:54:33 PM---Thanks! I'm pretty sure this only applies to service node pools however, and if "disjointdhcps" is s From: Dave Barry <[email protected]> To: xCAT Users Mailing list <[email protected]> Date: 10/11/2011 02:54 PM Subject: Re: [xcat-user] Why are service nodes sent commands for *all* nodes? Thanks! I'm pretty sure this only applies to service node pools however, and if "disjointdhcps" is set to 0. In this particular setup I have "disjointdhcps" set to 1: [root@mn ~]# tabdump site #key,value,comments,disable "blademaxp","64",, "domain","xcat",, "fsptimeout","0",, "installdir","/install",, "ipmimaxp","64",, "ipmiretries","3",, "ipmitimeout","2",, "consoleondemand","no",, "master","192.168.1.1",, "maxssh","8",, "ppcmaxp","64",, "ppcretry","3",, "ppctimeout","0",, "sharedtftp","0",, "SNsyncfiledir","/var/xcat/syncfiles",, "tftpdir","/tftpboot",, "xcatdport","3001",, "xcatiport","3002",, "xcatconfdir","/etc/xcat",, "timezone","America/New_York",, "useNmapfromMN","no",, "enableASMI","no",, "vsftp","y",, "nameservers","192.168.1.1",, "forwarders","4.2.2.1",, "installloc","storage01:/install",, "svloglocal","0",, "disjointdhcps","1",, "dhcpinterfaces","mn|eth1;service01|eth0;service02|eth0",, "dhcpsetup","n",, "dnshandler","bind",, So Service02 would not have any DHCP information for this node, yet my management node is still trying to talk to Service02. Is this a bug then perhaps? On Tue, Oct 11, 2011 at 1:26 PM, Linda Mellor <[email protected]> wrote: For the nodeset command in particular, often there is a single network between all service nodes and compute nodes. During compute node deployment, DHCP broadcasts can be picked up by any service node, so xCAT will set up identical DHCP configuration for ALL nodes on every service node so that any one of them can respond to the initial DHCP request, and then have the "next-server" value set to the correct service node for that compute node for subsequent deployment data requests. Linda Inactive hide details for Dave Barry ---10/11/2011 01:22:50 PM---Just curious, I noticed that even when a compute node is set tDave Barry ---10/11/2011 01:22:50 PM---Just curious, I noticed that even when a compute node is set to only use, for example, "service01" a From: Dave Barry <[email protected]> To: xCAT Users Mailing list < [email protected]> Date: 10/11/2011 01:22 PM Subject: [xcat-user] Why are service nodes sent commands for *all* nodes? Just curious, I noticed that even when a compute node is set to only use, for example, "service01" as their service node, the other service nodes still seem to be sent commands for that node. Example: [root@mn ~]# lsdef c1n01 Object name: c1n01 arch=x86_64 bmc=c1n01-bmc bmcport=0 chain=runcmd=standby currchain=boot currstate=netboot centos5.5-x86_64-compute groups=compute,all,ipmi initrd=xcat/netboot/centos5.5/x86_64/compute/initrd-stateless.gz installnic=eth0 interface=eth0 ip=192.168.1.2 kcmdline=imgurl=http://!myipfn!/install/netboot/centos5.5/x86_64/compute/rootimg.gz XCAT=!myipfn!:3001 ifname=eth0:00:50:56:11:11:15 netdev=eth0 kernel=xcat/netboot/centos5.5/x86_64/compute/kernel mac=00:50:56:11:11:15 mgt=ipmi netboot=pxe nfsserver=mn nodetype=osi ondiscover=nodediscover os=centos5.5 postbootscripts=otherpkgs postscripts=updaterepos.sh,syslog,remoteshell,syncfiles power=ipmi primarynic=eth0 profile=compute provmethod=netboot servicenode=service01 status=booted statustime=10-11-2011 12:13:27 [root@mn ~]# nodeset c1n01 netboot Error: Unable to dispatch hierarchical sub-command to service02:3001. This service node may be down or its xcatd daemon may not be responding. c1n01: netboot centos5.5-x86_64-compute As you can see, c1n01 has "service01" only as its service node, but the management node is still trying to talk to service02 about this command. Is this expected behavior, or have I misconfigured something? Thanks! ------------------------------------------------------------------------------ All the data continuously generated in your IT infrastructure contains a definitive record of customers, application performance, security threats, fraudulent activity and more. Splunk takes this data and makes sense of it. Business sense. IT sense. Common sense. http://p.sf.net/sfu/splunk-d2d-oct _______________________________________________ xCAT-user mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/xcat-user ------------------------------------------------------------------------------ All the data continuously generated in your IT infrastructure contains a definitive record of customers, application performance, security threats, fraudulent activity and more. Splunk takes this data and makes sense of it. Business sense. IT sense. Common sense. http://p.sf.net/sfu/splunk-d2d-oct _______________________________________________ xCAT-user mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/xcat-user ------------------------------------------------------------------------------ All the data continuously generated in your IT infrastructure contains a definitive record of customers, application performance, security threats, fraudulent activity and more. Splunk takes this data and makes sense of it. Business sense. IT sense. Common sense. http://p.sf.net/sfu/splunk-d2d-oct _______________________________________________ xCAT-user mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/xcat-user ------------------------------------------------------------------------------ All the data continuously generated in your IT infrastructure contains a definitive record of customers, application performance, security threats, fraudulent activity and more. Splunk takes this data and makes sense of it. Business sense. IT sense. Common sense. http://p.sf.net/sfu/splunk-d2d-oct _______________________________________________ xCAT-user mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/xcat-user ------------------------------------------------------------------------------ All the data continuously generated in your IT infrastructure contains a definitive record of customers, application performance, security threats, fraudulent activity and more. Splunk takes this data and makes sense of it. Business sense. IT sense. Common sense. http://p.sf.net/sfu/splunk-d2d-oct _______________________________________________ xCAT-user mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/xcat-user ------------------------------------------------------------------------------ All the data continuously generated in your IT infrastructure contains a definitive record of customers, application performance, security threats, fraudulent activity and more. Splunk takes this data and makes sense of it. Business sense. IT sense. Common sense. http://p.sf.net/sfu/splunk-d2d-oct _______________________________________________ xCAT-user mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/xcat-user ------------------------------------------------------------------------------ All the data continuously generated in your IT infrastructure contains a definitive record of customers, application performance, security threats, fraudulent activity and more. Splunk takes this data and makes sense of it. Business sense. IT sense. Common sense. http://p.sf.net/sfu/splunk-d2d-oct _______________________________________________ xCAT-user mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/xcat-user
<<inline: graycol.gif>>
------------------------------------------------------------------------------ All the data continuously generated in your IT infrastructure contains a definitive record of customers, application performance, security threats, fraudulent activity and more. Splunk takes this data and makes sense of it. Business sense. IT sense. Common sense. http://p.sf.net/sfu/splunk-d2d-oct
_______________________________________________ xCAT-user mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/xcat-user
