I have posted about this before, but here it is again. I would really like
Legato or someone else to shed some light on this issue.
Background:
When you add the client "foo.qbranch.se", the client name "foo" is
automaticaly added to the aliases list of the client. According to legato
documentation, legato suggests that each client should have both the full
(FQDN - Fully Qualified Domain Name) name AND the short name defined in
the aliases field.
What I want to do and what I have been doing:
In Networker 6, it works just fine to add the client "foo.qbranch.se" and
then edit the client after it has been creared and remove the short name
"foo" from the list of aliases. This works just as well in Networker 7.
HOWEVER, when upgrading windows clients to 7.1/7.1.1, recovers from the
networker user GUI don't work too well. For some reason, the client
reports the short name "foo" to the server, instead of the FQDN/long name,
which ofcourse means that the server can't find the client and says that
it is not a registered client. This is somethign new with networker 7!
Directed recoveries to the same client can still be done though, so we
have been living with this for a while now. The CLI command "recover"
works without any problems as well.
Today I made another interesting discovery. When using Networker user for
exchange, v4.0, the said program displays the error "client foo is not a
registered client", and when clicking OK the program immediately exits.
How am I supposed to recover anything exchange-related?? Adding "foo" as
an alias for the client "foo.qbranch.se" is not an option. See below for
details. As a side-note, backups of this client work just fine.
This kind of client definition works just fine for at least all windows
(NT/2000/XP/2003) and all UNIX clients we have (Linux, Solaris, OsX).
The question:
So, my question to Legato & co and others is: Why did this behaviour
change? Why can't one ONLY define the full-name of the client and live
without any further recovery-related problems?
It feels like Legato doesn't want to support using only FQDN-names, since
its not mentioned anywhere, and all examples explicitly state that you
should define the short name as well.
Why using short names is undesirable (read SUCKS):
If you, like we, backup several customers in a centralized environment,
you, as the hosting/service provider, don't really have a say about what
customers name their machines to. Thus, you get a lot of customers who
name their servers to very generic names, like "wwwserver" or "dbserver".
In order to backup two customers, who both have a server named
"wwwserver", Legato needs to be able to distinguish between those two
client definitions. Since the client name in legato is the unique
identifier of a client, each client name must be unique. Also, all aliases
must be unique. Thus, you need to define the client with the full/FQDN
name, and NOT list the alias/sort name "wwwserver" in the list of aliases,
since two clients can't share the same alias either.
For instance: wwwserver.customer-a.se must be defined with the full name,
if it should be able to coexist as a client on the same server which
wwwserver.customer-b.se is already defined. Also, both clients can't have
the alias "wwwserver" defined simultaneously either.
Summary:
In my opinion, using short names, is a legacy which doesn't scale well
either. Instead of defining the short name as well per default, the
supported/documented way of adding clients should be to only use the FQDN
name, unless the machine itself lacks awareness of which domain it belongs
to.
(And don't forget to answer my question if you have an answer - don't just
read the rant :P )
//Oscar
--
Note: To sign off this list, send a "signoff networker" command via email
to listserv AT listmail.temple DOT edu or visit the list's Web site at
http://listmail.temple.edu/archives/networker.html where you can
also view and post messages to the list.
=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=
|
