[3791] in Kerberos
Max. 'size' of Kerberos IV database
daemon@ATHENA.MIT.EDU (Mike Friedman (510) 642-1410)
Fri Sep 2 17:18:50 1994
To: kerberos@MIT.EDU
Date: Fri, 02 Sep 94 13:26:14 PDT
From: spgmnf@cmsa.Berkeley.EDU (Mike Friedman (510) 642-1410)
In order to test Kerberos V4 with a large database, I attempted to add 20000
principals (to a database that had only a couple of dozen entries). But
my expect script (that talks to kadmin) just stopped (with no messages in
any Kerberos log files) after adding principal number 11510. When I looked
at the 'virtual' size of principal.pag (with ls), I saw about 1.5 gigs! (The
'actual' size, shown with du, was 15M). I have plenty of file-system space
left, but I suspect I've about reached the maximum value of a file block
pointer for the sparse file.
I've been told by others that this is an inherent problem with ndbm, but, even
so, the virtual size I get with just 11K principals is about twice what at
least one other person got with 25K principals. Is this just a reflection of
my environment (SunOS 4.1.3_U1B)?
Is there any way around this limitation, that's reasonable to implement?
Using a substitute for ndbm might help, though if I could get even ndbm to
give me 25K or 30K principals, that would be an improvement. (As it happens,
unfortunately, when we eventually go production, we may have as many as 50K
principals).
Also, could someone at least tell me if V5 'solves' this, either because it
doesn't use (n)dbm, or for whatever reason?
Thanks.
*-------------------------------------------------------------------*
| Mike Friedman; Data Communication and Network Services; |
| 267 Evans Hall; University of California; Berkeley, CA 94720 |
| 1-510-642-1410 spgmnf@cmsa.Berkeley.EDU SPGMNF@UCBCMSA.BITNET |
| |
| If the law of gravity were repealed, Newton and Einstein would |
| turn over in their graves! |
*-------------------------------------------------------------------*
