How To:Diagnose "host down" errors

From wiki.zmanda.com
Jump to navigation Jump to search

This article is a part of the How Tos collection.

By John R. Jackson, Technical Software Specialist, [email protected].

Check each and every item below very carefully, even if you think you've already done so. I can't count the number of times I've posted this list, had someone say they checked it all then found out later an item on it really was the problem.

And don't despair. For whatever reason, this one area (getting a client configured) causes **everyone** grief until they've done it a few times. It would be nice if it could be automated, but OS's are so different in these areas it's next to impossible.

Checklist

Are there any system log messages from (x)inetd about amanda or amandad?

Are the Amanda services in /etc/services or your NIS map?

If you use inetd: Is there a line in /etc/inetd.conf to start amandad?

If you use xinetd: Is there a file named /etc/xinetd.d/amanda?

If you just added the line to /etc/inetd.conf, or edited /etc/xinetd.d/amanda, did you send a HUP signal to (x)inetd?

Looking, very, very closely at the inetd.conf line, respectively the content of /etc/xinetd.d/amanda, make sure:

- it's the right service name

- the service name is registered and has the same number everyplace

- the protocol is udp, not tcp

- the path to amandad is correct

- the user to run amandad as is correct

- the user to run amandad as can get at the amandad binary and all the associated shared libraries (e.g. no locked up directories in the path)

- there are no extra or missing fields

- the line is not broken into two lines (e.g. a bad copy/paste)

If you have lsof (ftp://vic.cc.purdue.edu/pub/tools/unix/lsof/), make sure inetd is listening on the Amanda port. Or you can use:

     $ netstat -a | grep -i amanda
         *.amanda                              Idle

If you are using a firewall or TCP wrappers, make sure it is set up to allow the amanda service in to the client from the server.

Is /tmp/amanda/amandad*debug getting updated?

Is the access time on the amandad executable getting updated (ls -lu)?

If you run the amandad program by hand as the Amanda user on the client, what happens? It should sit for about 30 seconds and then terminate.

files in /etc/xinetd.d

Refer to these examples and ADJUST THEM to your user/group/paths ....

#/etc/xinetd.d/amanda
service amanda { 
       disable = no
       socket_type             = dgram
       protocol                = udp
       wait                    = yes
       user                    = amanda
       group                   = disk
       server                  = /usr/local/libexec/amandad
}
#/etc/xinetd.d/amandaidx
service amandaidx { 
       socket_type             = stream
       protocol                = tcp
       wait                    = no
       user                    = amanda
       groups                  = yes
       server                  = /usr/local/libexec/amindexd
}
#/etc/xinetd.d/amidxtape
service amidxtape { 
       socket_type             = stream
       protocol                = tcp
       wait                    = no
       user                    = amanda
       groups                  = yes
       server                  = /usr/local/libexec/amidxtaped
       server_args             = amidxtaped
}