SUMMARY: Red Hat Linux 8.0 hangs as an LDAP client when LDAPserver goes down
Greetings, again.
I seem to have found a working solution for our original woes.
Lans Carstensen suggested that we lower the timeout for the NSS resolver
library, NSS_LDAP. He also suggested that we make the LDAP PAM module
return on system errors. Here's what he suggested, in his words:
---------
First off, upgrade to the latest/greatest nss_ldap from ftp.padl.com and
configure bind_timelimit to something small like 10 seconds. The
default I believe is 15 minutes or somesuch, which means 15 minutes to
time out per process creation. It's absolutely insane, and that's why
you're having to reboot systems - they're stuck in the resolver library
waiting on a huge timeout.
Secondly, pull the line from /etc/pam.d/system-auth that looks like:
account [default=ok user_unknown=ignore service_err=ignore
system_err=ignore] /lib/security/pam_ldap.so
I believe that's what prevents root logins. My read on it is that it's
quashing errors being returned in pam_stack and is therefore not letting
the stack rules fall through properly to allow a local root login. I
could be wrong, but it works for us.
Finally, if you're serious about LDAP I'd recommend getting a support
contract with Luke Howard's company, PADL software. (http://www.padl.com/)
---------
I'm beginning to think that this configurable behavior should be the
default. I'm CCing this message to the Core development mailing list at
open-it, where we discuss these issues. I'd also suggest extending
NSS_LDAP to create a persistent local cache, which should only be used
when the LDAP server is down. This would drive LDAP authentication one
step closer to hands-free offline authentication (a worthy goal). I'm
sure the path to this goal is filled with thorny issues, but I'm also
sure it's worth the job. Especially for corporate customers =).
I'm shamelessly plugging Directory administrator
(http://diradmin.open-it.org/) as a user/group management tool, for
those of us who are using LDAP authentication. It still has a way to
go, but I plan to build several enhancements to improve large-scale
deployment scenarios.
luck,
Manuel Amador
_______________________________________________
LinuxManagers mailing list - http://www.linuxmanagers.org
submissions: LinuxManagers@linuxmanagers.org
subscribe/unsubscribe: http://www.linuxmanagers.org/mailman/listinfo/linuxmanagers
[Home]
[Kernel List]
[Linux SCSI]
[Video 4 Linux]
[Linux Admin]
[Yosemite News]
[Motherboards]