savefs: unknown host name

[pscherbr]

I'm hoping that someone can shed some light on this strange problem I've run into with Sun's Solstice Backup/Networker.

I have a server named 'toolbox' running Solaris 8 & Networker 6.1.4. I also have a client named 'supernova' running Solaris 9 & Networker 6.1.4. The server currently backs up 14 other machines, so I'm reasonably confident that the server is working correctly.

When I attempt to backup the client from the server via the nwadmin GUI, the backup fails with the following error message:

savefs: unknown host name: toolbox

According to the documentation this means that the client can't resolve the hostname 'toolbox'. However, the client can resolve the hostname 'toolbox'.

Here's my /etc/hosts file from the client machine (I've obscured the IP and modified the domain name for security):

# cat /etc/hosts
xxx.xxx.xxx.4 supernova supernova.domain.com loghost
xxx.xxx.xxx.35 toolbox toolbox.domain.com

And my nsr/res/servers file:

# cat /usr/local/nsr/res/servers
toolbox
toolbox.domain.com

I can also ping the server from the client using both the short and long names:

# ping toolbox
toolbox is alive

# ping toolbox.domain.com
toolbox.domain.com is alive

Finally, when I attempt to initiate a backup from the client machine, it only works if I do not specify the server. (I've deliberately used the -n flag for the 'save' command here, but it does backup successfully without it.)

# save -c supernova -g Supernova -n -q -s toolbox /var/
save: unknown host name: toolbox

# save -c supernova -g Supernova -n -q -s toolbox.domain.com /var/
save: unknown host name: toolbox.domain.com

# save -c supernova -g Supernova -n -q /var/
save: /var 4594 records 896 KB header 14 GB data
save: /var 14 GB estimated

I've also tried restarting the Networker software on both the server and client machines. I'm starting to run low on ideas. I could run a cron job to manually backup the client every night, but I would much rather see if I can fix this problem.

Thanks for your help.

Peter

 

[605]

Ping name does not always help - it will not let you know if the name resolution is wrong. And this is what seems to cause the problem. This may happen if you use a name service as well.

Try to use the ip address instead of the name - this should work as well. If it does, name resolution is definitive the problem.

 

[pscherbr]

Hmm... I just tried it using the IP address and it still doesn't seem to work.

# save -c supernova -g Supernova -n -q -s xxx.xxx.xxx.35 /var/
save: unknown host name: xxx.xxx.xxx.35

So then I edited my nsr/res/servers file, added the IP address to the end of the file:

# cat /usr/local/nsr/res/servers
toolbox
toolbox.domain.com
xxx.xxx.xxx.35

Restarted the Networker daemon, and tried again:

# save -c supernova -g Supernova -n -q -s xxx.xxx.xxx.35 /var/
save: unknown host name: xxx.xxx.xxx.35

Is there somewhere else I need to enter the IP address of the server?

 

[605]

This is weird. What happens if you login to your client and do the following:

# cd /var
# save -s servername [or ip]

Also please make sure that nsrexecd on the client is also up. I have heard that Legato changed this recently for at least 7.x

 

[pscherbr]

605,

Still no luck.

# cd /var/

# save -s toolbox
save: unknown host name: toolbox

# save -s toolbox.domain.com
save: unknown host name: toolbox.domain.com

# save -s xxx.xxx.xxx.4
save: unknown host name: xxx.xxx.xxx.4

# save -g Supernova
save: Using toolbox as server
<Output deleted>

(I had to use the -g flag since otherwise the backup wanted to use a tape from the Default pool, and I don't have any tapes assigned to that pool.)

# ps -ef | grep nsrexecd
root 17839 17838 0 12:30:10 ? 0:00 /usr/sbin/nsrexecd
root 17838 1 0 12:30:10 ? 0:00 /usr/sbin/nsrexecd

I think I'll try reinstalling Netwoker 6.1.4 to see if that helps any. Otherwise I may try installing 6.1.3 since I have other client machines working just fine with that version. This is the first client I tried using the 6.1.4 version on.

Thanks for your help.

Peter

 

[kcongdon]

Can the client resolve the Legato server's name? Does the client have problems resolving any other hostnames?

Is there a DNS server in the environment? If so, is either the client or server configured to use it instead of a hosts file? Can either client or server be configured to use a DNS server? What would the nslookups by name and IP address from both sides return?

 

[605]

(I had to use the -g flag since otherwise the backup wanted to use a tape from the Default pool, and I don't have any tapes assigned to that pool.)

This maybe true. I think the better solution is to run "save -s server -b pool ..." - this will override the automatic data sorting routine and write data to the pool you specify here.

 

[wallace88]

Start back to the basics... test the network connectivity first. Refer to LEGATO TECHNICAL BULLETIN 299: IP Naming in Heterogeneous Environments (UNIX | NT | NetWare | Windows 95)

http://www.legato.com/resources/bulletins/299.html

Now, on the machine supernova:

You have a host file setup on the client supernova, BUT, this is only used if your /etc/nsswitch.conf file allows Solaris to look at the host file. Edit the /etc/nsswitch.conf and look for a line that starts with the word "files". For example:

hosts: files dns [notfound=return] nis

The above example tells Solaris to resolve hostnames by the host file first before looking at the dns tables, then the NIS tables.

Once you have set up the nsswitch.conf to look at your host file, then refer to the above technical bulletin to verify network connectivity. Try:

ping -s toolbox
ping -s toolbox.domain.com
ping -s (toolbox's ip address)

Make sure that ping returns the correct host name, fqdn, and ip address for toolbox.

Next: run ls -ld /etc/hosts. Make sure that file permissions allow anyone to read the file.

Try the following:

grep -i toolbox /etc/hosts

How many lines does it return?


Stop NetWorker and delete the directory /nsr/tmp. hostname/ip information is cached in this directory. When you restart NetWorker, the directory will be recreated.


Just curious... in your above information, when you specified the ip address in the ping and save commands... did you actually put the ip address (such as 10.1.1.4)? or did you literally put: xxx.xxx.xxx.4

 

[eliellis]

I had the same problem. What happened was another SA changed the the host entry in /etc/hosts (ie changed the host entry for supernova in supernova's hosts file). In your environment it would have looked like this:

Original /etc/hosts entry on supernova:
xxx.xxx.xxx.4 supernova supernova.domain.com loghost

/etc/hosts entry when error began to occur on supernova:
xxx.xxx.xxx.4 someOtherName supernova supernova.domain.com loghost

If it never worked in the first place, be sure that the name you have specified in the NetWorker gui is the same as the name that is directly after the IP address in /etc/hosts.