Dear Mark,

> On a more detailed look at this the bug is in fact based on a
> misunderstanding of the purpose of ypxfrd. Rather than accepting and

You are right. I am very embarrassed to realize that both bugs I reported on 
your package are invalid.

I am so sorry for wasting your time on this.

Re: bug #186363: By a strange feat of misfortune, the debugging I did on my 
slave server consistently lead me to the wrong conclusion. I was too fast in 
reporting to LP, and not careful enough in checking.

When I started ypxfrd on the slave, the map transfer no longer hung on the 
master, but that was coincidental. That led me to conclude that the daemon 
must run there. It has been app. 10 years since I set up the NIS system, and 
I have forgotten about the details in the meantime. You are right that ypxfrd 
should run on the master, _not_ on the slaves.

As for the other one, bug #186363, it is also invalid.

What happened was that our domainname recently changed, and it  seems that the 
old master name remained embedded in those databases that are not updated 
often, and the new master name was embedded in the rest. Consequently, when 
the slaves run ypinit, they only received maps associated with the new master 
name

I was not aware that the database entries identified their master, but 
thinking about it, it is absolutely reasonable.

What I _should_ have done -- and have now -- was to force the master to update 
all maps from the "source files" when noticing that the slaves did not 
enumerate all the master's maps.

It is possible to provide a script that would reveal the situation. One 
solution is to see how many hosts are listed in the second field when doing 
ypwhich -m. For example:

  ypwhich -m|cut -f2 -d' '

should only list one master host. Also, the output of:

  ypwhich -m|cut -f1 -d' '|sort

and

  /usr/lib/yp/yphelper --maps <master> |sort

should be identical. If they are not, it indicates that the master's maps 
should be updated from source files. (The yphelper command is very finicky on 
what hostname it is given, in my case I need to use the FQDN.)

So, the only thing that has come out of this is that the commented-out line in 
ypinit, that has $MASTER$, should be $MASTER. A bug in a comment line. 
Wow... :-(

Cheers,
Morten

-- 
nis init.d script should start ypxfrd on slave, not master
https://bugs.launchpad.net/bugs/186363
You received this bug notification because you are a member of Ubuntu
Bugs, which is the bug contact for Ubuntu.

-- 
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Reply via email to