V100 boot hangs on NIS

Question

Hi, I've got a V100 server at my company that has Solaris 8 installed on it (I now, very old). One of our staff tried to install LDAP user authentication on our domain and set this server to be a client. This all went very badly and I'm trying to recover the system, the problem is that this server now hangs at boot very early on at the domain resolution. when I reset the machine the boot hangs at the below output
checking ufs filesystems
/dev/rdsk/c0t2d0s6: is stable.
4/4/2005 5:55:4 GMT LOM time reference
NIS domainname is ingena.localHow can I fix this? would like to change the nsswitch.conf file so it stops looking for LDAP and NIS servers but I can't get onto the disk to edit this. Any ideas???

Answer 1

Having just had a grand old time (not!) with deploying Directory Server (my 64 page installation and troubleshooting manual will be forthcoming very soon), I believe (but I won't say that I'm certain) that your problem is one of a few things, all of which can build on each other:
* The client cannot connect to the directory server and is sitting there waiting for a response
* You are attempting to resolve root through LDAP, as in the system is looking to LDAP first instead of local /etc files (bad idea!)
* Your pam.conf settings are incorrect and thus cannot resolve the "root" user, thereby making it impossible to continue to run the startup scripts
* Your nsswitch.conf file is configured to communicate through LDAP first instead of the appropriate order causing the inability to resolve host names properly. (See below.)
As to the NIS domainname issue, that only means that it found an entry in /etc/defaultdomain. That doesn't necessarily mean that you're still running NIS. In fact, chances are that you're not utilizing NIS any more. When you initialize an LDAP server, the ldapclient command makes a very arrogant and stupid assumtion that every bloody thing in /etc/nsswitch.conf will somehow be resolved through LDAP - netmasks, bootparams, ethers, hosts, etc.
So, follow these guidelines:
* Keep root local. Do not rely on LDAP. If you store root in LDAP, (A) the client can't boot if the LDAP server is down because root cannot be resolved and (B) anyone who breaks into the LDAP server could get the keys to all of your castles.
* As soon as you initialize a client, change the /etc/nsswitch.conf file back to what it should be. Too many functions (like NTP) break unless you import ALL of your relevant /etc files into LDAP, which is a very, very, VERY stupid assumption for Sun to make.
Regardless, boot into single-user mode and look at the nsswitch.conf file. If that doesn't give anything really viable, run each startup script individually to see which one is causing the problem.