Re: [Testbed-admins] Random DNS problems on boss, can't resolve tips or test pc's

[Leigh Stoller <stoller@flux.utah.edu>]

> We've been having several problems getting new nodes to correctly
> load our STD OS image (amongst other problems)...  but one problem
> that seams to be constantly plagueing us right now is that the DNS
> server on boss isn't getting updated correctly or is failing
> someplace.  At random times, DNS will fail for tips and our test
> machines (pc1, pc2, etc.)  it always works fine for boss, ops, and
> our mail server (all these are in the hosts file) but the dynamic
> machines have been in constant flux it seams like over the past few
> days.  I suspect it is because we keep trying to get machines to
> boot and load an image, we are moving machines between HWDOWN and
> reloading projects, which I beleive changes the DNS names of the
> machines.  but I don't think the DNs update app is completeing, or
> it isn't being run or something.

Hi Ben. named startup errors land in /var/log/messages, so check
that file to make sure that named is starting okay. Most errors cause
named to give up and die.

The interesting thing is the tip machines; they should be hardwired
into your map files in /etc/namedb along with the entries for boss,
ops, etc.

When things do fail, check /etc/namedb/*.db to make sure the pcs are
listed there, and also check /var/log/messages for errors.

Also, named_setup runs only during experiment setup and teardown. If
you are moving nodes between hwdown and reloading by hand, you need to
run named_setup by hand as well. But that would not affect the tip
machines, except if named had died.

Lbs