Hello,
I only got two responses to my query (I think -- others may have been
lost when my new e-mail account was not yet behaving properly). Thanks
to
Satish Movva
Jason Marshall
for taking time out of their busy day to respond.
Satish indicated that he had similar problems with get_netmask, and a
re-installation of the jumpstart directory (preserving the config files)
fixed the problem for him.
Jason indicated that non-default netmasks might cause this (ie. not on
an octet boundary), but that is not our case here. All subnets are
Class C subnets. Jason did suggest setting the correct netmask in the
bootparams file. I didn't explore this option because the bug reports
said that this work around would not work
Thanks again to both for their ideas, but I guess we'll stick with our
current work around.
If you have any other thoughts on this, please let me know. Otherwise,
I'm happy that this summary will exist in the archives so the next poor
soul who does a search for "get_netmask" doesn't draw a blank.
Take care,
Mike van der Velden
Siara Research Canada, Inc.
mvanderv@siara.com
Original Question:
>
> Hello,
>
> I recently joined this company, and the first thing I was asked to do
> was debug the jumpstart installations.
>
> Jumpstart was working fine for months. All client systems were in a
> single subnet, and the boot server, install server , and NIS server were
> the same box (Ultra 10, running 2.6 5/98). Recently, a NetApp 730 was
> installed as the new fileserver, and it is now acting as the install
> server. The Ultra 10 is still the boot server and NIS server.
>
> Ever since the move to the NetApp, however, jumpstart installations hang
> at the point where the console reads, "Configuring devices..."
>
> We checked and doubled checked our jumpstart configuration, making sure
> rarpd, tftpd, were working, making sure that bootparams, ethers, hosts,
> netmasks, dns, were up to date, and that the proper filesystems were
> exported with read-only permissions. The nsswitch.conf file on the
> servers was configured to look at local files only, but we added the
> info to NIS as well. The rules and sysidcfg files had not been changed
> in recent memory. Nothing seemed to help.
>
> In the course of debugging, I found the rcS script in the boot
> filesystem, which echoes the string "Configuring devices..." I turned
> on verbose mode ("set -v" ) in the script, and noticed that rcS actually
> hangs at the part where it runs a program called /sbin/get_netmask.
>
> The Sun Managers archives came up blank on this problem, but SunSolve
> came up with 4 hits on get_netmask problems during jumpstart. Bug
> reports only, but no patches. The problem has been around since Solaris
> 2.5, and also appears in Solaris 7. Here are the bug id's for those of
> you who are curious: 40123135, 4030522, 4218092, 4251034.
>
> It looks like get_netmask is sending out an ICMP Address Mask Request to
> the server, but never gets a response. One workaround suggested is to
> ping the workstation making the request. While this works quite nicely,
> it does mean that jumpstart is no longer a hands-off affair. The
> workaround we adopted was to comment out the get_netmask call in rcS,
> and hard code our netmask instead. This is ok for our network, but
> cannot be acceptable for larger sites. Further, we'll have to remember
> to make the same fixes to any new OS rev's we wish to deploy via
> jumpstart.
>
> I have two questions, the first of which is rhetorical.
>
> 1. Why hasn't Sun fixed a bug like this that has been around since 2.5?
> It sure cost our site a lot of time, not only debugging a known bug, but
> also in the delays in deploying new workstations.
>
> 2. Has anyone else experienced this problem? Have you found a proper
> fix, or what other workarounds have you used?
>
> Mike van der Velden
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:14:03 CDT