"lease imbalance - lts" flood - please help

[David W. Hankins]

We've seen this problem from several viewpoints now...it took awhile
to figure out what was being reported (because the log message itself
is a false negative).

I think what happened is in 3.0.x somewhere we let the servers
check lease balance state upon every attempt to allocate addresses,
not merely on every successful attempt.

The rationale was that if a pool goes totally slantwise - where
let us imagine the primary has a hundred free leases and the secondary
has zero backup leases - where the server can not allocate a lease,
then it will not check balance, so rebalancing will never happen...and
the servers languish in this state, one server incapable of giving
leases and the other happily oblivious to any problem.

On busy servers, however, the consequence here is that the server
will get overloaded scanning the pools to rebalance...the secondary
'spams' the primary with POOLREQ messages.

It's an insult to injury, then, that the log message emits more often
than it should (for some rather complicated sausage-making reasons)
and isn't helpful in diagnosing anything.

On Thu, Aug 30, 2007 at 04:19:58PM -0700, ashley.hatch at unlv.edu wrote:
> I have found several others who had this problem, but I have yet to find 
> anyone come up with the solution, at least that they reported back to the 
> newsgroup.

For 3.0.x, we probably need a source change; refuse to process
POOLREQs too frequently or somesuch.

3.1.x processes pool rebalance events on a schedule, rather than
driven from events, so running 3.1.0 is an effective workaround for
the 'spammed with lease imbalance messages' problem.

This is assuming that there isn't a lease database inconsistency (a
bug by another name...) or that 'faulting' the database on one
server to make both consistent isn't helping.  I mention this largely
for the archive's benefit, since you've said so.

-- 
Ash bugud-gul durbatuluk agh burzum-ishi krimpatul.
Why settle for the lesser evil?	 https://secure.isc.org/store/t-shirt/
-- 
David W. Hankins	"If you don't do it right the first time,
Software Engineer		     you'll just have to do it again."
Internet Systems Consortium, Inc.		-- Jack T. Hankins