Move of DNS Server

[MarkSyder]

Hi everybody.

I've decided it's time to move DNS off our out-of-support 10.20 server on to a server running 11.11.

I made the move last week and only stopped named from running on the old server after the new system had been running smoothly for several days. To my surprise, all the workstations started reporting loss of NFS mounts.

I've confirmed that the /etc/resolv.conf file shows the new server and its backup, and does not contain any information relating to the old DNS server. Once I restarted named on the old server, the NFS mounts were restored.

I suspect there must be a file somewhere on the workstations that still points to the old server, but I've no idea where it is or what it's called. Any ideas?

Mark Syder (like the drink but spelt different)

[RAC_1]

Does all nfs clients resolve new and old ip of dns server? They should be pointing to new one. How do you use nfs clients? Manual nfs mounts? automount? what?

You may have to restart nfs clinets if nsllokup on ips/names is not correct.

[MarkSyder]

Thanks RAC.

Interestingly enough, the NFS mounts that went down were on the new DNS server and its backup.

NFS mounts are done by automount.

All workstations are correctly resolving DNS information for the new DNS server, its backup, and the old one. The output from nslookup indicates that they are looking at the new server.

Mark

[RAC_1]

Were the automounters restarted between dns change?

[Chan 007]

Mark,

I suspect your /etc/exports file on your NFS server which may contain the old domain name.

Can you check that

Chan

[MarkSyder]

No, the automounters weren't changed, but I did reboot one workstation to no avail (only one person had reported the fault, and I thought it was local to his machine).

The domain name didn't change, just the DNS server.

I've just amended the hosts file to contain the new DNS server and the nsswitch.conf file to point to files first then DNS. Is this likely to help?

Mark

[Chan 007]

Mark,

the nsswitch.conf should have dns and then files.

hosts: dns files


Chan

[MarkSyder]

Chan,

See Patrick's answer in http://forums1.itrc.hp.com/service/forums/bizsupport/questionanswer.do?threadId=959810

I've read a lot of Patrick's answers, and if he says files first, I believe him!

Mark

[RAC_1]

Should files be before dns depends on what you want. If you have just few hosts to manage, then I would put hosts in file-/etc/hosts and it is lot quicker than using dns. So answer is it depends. To isolate the problem, try putting all required ips in /etc/hosts first and check. Then put all those ips in dns and check. Is the new dns server as primary dns server for all your clients? If automounter is in play, what do you use? direct/indirect maps? How does resilution happens for maps?

[MarkSyder]

Interesting reply, RAC.

It's a fairly small network, so I think I'm ok putting essential servers in each /etc/hosts file.

Yes, the new DNS server is the primary DNS server for all my clients.

How do I know if the maps are direct or indirect?

Mark

[RAC_1]

What does automount directive in /etc/nsswitch.conf say? There is absolutely no harm in putting al ips in /etc/hosts file.

[MarkSyder]

automount: nis [NOTFOUND=continue UNAVAIL=continue] files

Mark

[rick jones · Accepted solution]

IIRC a process only reads /etc/resolv.conf once and then caches the results of that read. Soooo, if you edit /etc/resolv.conf while a long-lived process remains running, long-lived processes will still have the old DNS server information. Only new processes get the new DNS server information.

The "easiest" way to get all new processes is to reboot the systems in question.

[MarkSyder]

Thanks Rick.

I actually did reboot the workstation of the first person who reported it, but the fault persisted.

A colleague tells me he has encountered this before and had to flush the routing table (it's a long time ago and he's unsure how he did it). I've had a look at the man page for route and it appears I need to run route -f.

Anything I need to know before I run this?

Mark

[RAC_1]

Reboot should have taken care of it. Flushing a routing table is not a problem.But I doubt, how it is different than reboot? (Unless path to new dns server is changed.)

[RAC_1]

Reboot should have taken care of it. Flushing a routing table is not a problem.But I doubt, how it is different than reboot? (Unless path to new dns server is changed.)

Another thing to lookout is as follows. Since you have problem with nfs clients mounting the shares (and you are using automounter for it), you can send -usr2 signal to automuter and it will start logging debugging messages in /var/adm/ directory. Looking at log file may give you some information to start with.