Subject: 3.0 YP lookup latency
To: None <firstname.lastname@example.org>
From: Stephen M. Jones <email@example.com>
Date: 06/20/2006 13:08:53
I'm working out a migration to 3.0 and have setup a client machine for
testing with the 3.0 base release from December and a new kernel from
the May 27th source build.
I've tried setting up the 3.0 host as a yp client and a slave from the
master, but I'm still seeing a 15 or so second latency on uid mapping.
However, ypcat requests return as quickly as expected. The ypserver and
slave are both NetBSD 2.1, but I'm not sure how this could be an issue.
So I turned on logging on the ypserver and discovered that any yp lookup
(using id, whoami, sudo et cetera) would trigger a next_2: request for
every single user id in the password file. This explains the latency.
Is this a bug or an incompatibility? I won't be able to upgrade my
ypserver and ypslave for a couple of months, and if its an incompatibility
I would assume it would be resolved once everyone is happily running
3.0. I actually tried copying over the 'ypbind' from the 2.1 system to
the 3.0 and restarting it, but I get the same results.
$ time id smj
load: 0.35 cmd: id 16846 [poll] 0.12u 0.20s 0% 1472k
load: 0.35 cmd: id 16846 [poll] 0.24u 0.41s 2% 1472k
load: 0.35 cmd: id 16846 [poll] 0.43u 0.69s 4% 1472k
load: 0.32 cmd: id 16846 [poll] 0.53u 0.84s 4% 1472k
load: 0.32 cmd: id 16846 [poll] 0.58u 0.93s 5% 1472k
uid=1001(smj) gid=0(wheel) groups=0(wheel),500(arpa),800(instructor),600(MetaARPA)
16.16s real 1.65s user 2.65s system