SUMMARY: panic - boot: mismatched version of /boot interface: new boot, old kernel

This note is an FYI / Summary of a jumpstart problem into which
I ran.  It is offered here in the hopes that someone else who
is battering his/her head against the proverbial jumpstart-troubleshooting
wall might find this in the archives and thus limit their 
head smashing...

Problem:  Jumpstart of a newly added system fails with:

      panic - boot: mismatched version of /boot interface: new boot, old kernel


Solution:  I had duplicate IPs in my /etc/hosts file:  changing the IP of the
	   new system (an E3000) resolved the problem.

	   This error occurs when the boot file /tftpboot/<IP Addr in Hex> does
	   not match the 'root' entry in the /etc/bootparams file.  This mismatch
	   may be caused by:

	   	1) an incorrect inetboot file or link in /tftpboot
		2) an incorrect /etc/bootparams entry
		3) the wrong /etc/bootparams entry being used by tftpboot

	   I am guessing that most times these problems are probably caused by
	   either not using the 'add_install_client' script or running 
	   'add_install_client' incorrectly.  Problems can occur when

	   	o /etc/hosts (or other nameservice) is updated incorrectly
		o /etc/ethers is updated incorrectly
		o /etc/bootparams is updated incorrectly
		o /tftpboot directory is updated incorrectly
		    - the file <Host IP in Hex> is linked to the wrong inetboot file
		    - inetboot file was never updated

		NOTE that 'updated incorrectly' can mean typos, or, duplicate 
		entries.

	    Manually confirm that /etc/hosts, /etc/ethers, and /etc/bootparams
	    entries are correct and *unique*.

	    In my case I had inadvertantly (and stupidly) picked an IP address
	    that already existed in the /etc/hosts file.  So, 'rpc.bootparamd' looked
	    up IP address 192.168.1.6 in /etc/hosts and got 'huey' instead of
	    'e3000', and subsequently looked up 'huey' in /etc/bootparams
	    instead of 'e3000'.  'huey' is an older entry for a non-existant
	    system (well, it is not up at the present time) that had a root
	    entry of 

	      root=aslan-hme1:/export/install/2.6/5-98/Solaris_2.6/Tools/Boot

	    but 'e3000' was using a Solaris 8 (07_01) inetboot file:

	    % ls -l /tftpboot/C0A80106
	    <snip> ... <snip> /tftpboot/C0A80106 -> inetboot.SUN4U.Solaris_8-2

	    so 'e3000' was booting Solaris 8, but was then sent to a Solaris 2.6
	    root directory... not so good.



Troubleshooting steps:

	1) boot -v net - install

		NOTE:  this step revealed the problem, but I was too dense
		       to see it at this point.

		       'boot -v' is very helpful at giving additional boot information.
		       The output can be very clarifying, especially useful in diagnosing
		       jumpstart problems.

	2) sunsolve search on 
	
	   "boot: mismatched version of /boot interface: new boot, old kernel"

	   turned up BugID 4167232, "Solaris 2.5.1 net boots not supported anymore",
	   but I have neither an E10K nor Solaris 2.5.1, so this was useless.

	3)  Google search on "mismatched version of /boot interface"

	    very helpful!  found the Sunmanagers archives that have previously discussed
	    this topic (in particular, *not* using 'add_install_client', but rather
	    editing files manually)

	4)  Manually inspecting /etc/hosts, /etc/ethers, /etc/bootparams


I hope I wrote clearly, and I hope this message might help someone in the 
future.

--
Steve Camp
Camp Technologies, LLC
steve@camp.com
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers