Regarding "genesis" functionality: If you have a dynamic range defined in your provisioning network (dynamic range field of networks table), then xCAT will try to discover anything that PXE boots on that network. This assumes you have site.dhcpinterfaces defined to limit DHCP to the provisioning net. To disable automatic discovery, remove the dynamic range, and run "makedhcp -n".
If you are trying to get recently installed cluster nodes to not PXE, the easiest way is modifying the bootorder. If you are using UEFI install, the OS should do this for you by inserting its own boot entry at the top of the list (e.g., "Red Hat Enterprise Linux"). Once that entry is there, the node should not be PXE booting unless the boot entry is corrupted. If you're using legacy boot, you can put the HDD before the network (pasu <node> set BootOrder.BootOrder "Legacy Only=Hard Disk=PXE Network"). In either case, if you want to reinstall, you can use the rsetboot command (or nodeboot in confluent) to PXE on the next reboot. If recently installed nodes are still being discovered, my guess would be that the MAC is not discovered. A mac might be discovered, but notnecessarily the one it's trying to boot from. You can enter multiple macs for a node and specify NOIP for macs you don't want it to boot from (mac table entry: node,,"1:2:3:4|5:6:7:8!NOIP"). Make sure to run makedhcp after doing this. If all this doesn't work, what does "nodeset <node> status" return? Perhaps the node's install status is not getting updated correctly. Regarding the ssh permissions, I'm not sure. It looks like the remoteshell postscript is slated to run. Perhaps check /var/log/xcat/ on the node for postscript logs to get a hint? What does /install/postscripts/_ssh look like on the management node? How about root's .ssh dir also on the management node? Regarding otherpkgs: The otherpkgs postscript should be under /install/postscripts with all postscripts. Things to remember when using otherpkgs: The otherpkgdir is a repo. Make sure you run createrepo on the dir after adding packages. The otherpkglist is a list of package names, not file names. So you would want "ssh-server" rather than "ssh-server-1.2.3-4.el7.x86_64.rpm". If you want the 32-bit version of a package, you can specify "ssh-server.i686" for example. Regards, Christian Caruthers Lenovo Professional Services Mobile: 757-289-9872 -----Original Message----- From: James Ault via xCAT-user <xcat-user@lists.sourceforge.net> Sent: Sunday, August 9, 2020 23:50 To: xcat-user@lists.sourceforge.net Cc: James Ault <aul...@yahoo.com> Subject: [External] [xcat-user] XCAT 2.16 Discovery Mode - Genesis Hello. I am running XCAT 2.16 on CentOS 7 on an air-gapped network and I have a few questions: 1) how may I disable the "genesis" discovery functionality? A previous node management system had a default behavior of booting from network, and only the nodes that were marked for install would be installed, but all other nodes would boot from local disk with a slight delay. Many of these nodes are getting stuck in a boot loop where they boot genesis instead of their own OS from local disk, and the genesis process never completes, it just loops forever. I need to make this discovery process work properly or shut it off completely. One relevant error message on console: "xcat.genesis.minixcatd: The request is already processed by xCAT master, but not matched." (repeat every 5 minutes for ever) 2) I have successfully installed an OS (CentOS 7.8) on a few nodes, but the nodes somehow are configured with ssh key files that have permissions "640" which cause the sshd to fail and exit with an error, which means I cannot login remotely. This does not seem like a reasonable default. If there is a configuration setting in XCAT that will help me fix this before the OS install is finished that would be very helpful. 3) I want to install other packages during the post install process, but any attempts so far have not succeeded: Example typed from printed logs: xcatmn# lsdef node1 node1: arch=x86_64 bmc=10.10.10.10 bmcpassword=(enter_password_here) cons=ipmi consoleenabled=1 groups=all,x86_64 mac=1:2:3:4:5:6 mgt=ipmi netboot=xnba os=centos7.8 postbootscripts=otherpkgs postscripts=systlog,remoteshell,syncfiles profile=compute provmethod=centos7.8-x86_64-install-compute routenames=defaultroute status=failed status=(insert date here) xcatmn# lsdef -t osimage centos7.8-x86_64-install-compute imagetype=linux osarch=x86_64 osdistroname=centos7.8-x86_64 osname=Linux osvers=centos7.8 otherpkgdir=/install/post/otherpkgs/centos7.8/x86_64 otherpkglist=/install/custom/pkglist/compute.otherpkglist.txt partitionfile=/install/custom/partition/compute-default-partition.txt pkgdir=/install/centos7.8/x86_64 pkglist=/opt/xcat/share/xcat/install/centos/compute.centos7.pkglist profile=compute provmethod=install template=/opt/xcat/share/xcat/install/centos/compute.centos7.tmpl I cannot find the "otherpkgs" script. I attempted to use "chdef" to create a "otherpkgs" attribute on the osimage, but it gave an error saying that attribute was only valid for an OS type of "NIM" or something that did not match my environment. I am trying to install 'ansible" and all the dependent packages have been gathered and copied to the "otherpkgdir" above and the list of packages in that dir is included in the "otherpkglist". Sorry if I made typos above, and thanks in advance for your help. -Jim _______________________________________________ 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
