Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!utgpu!water!watnot!watmath!clyde!cbatt!ucbvax!tekchips.tek.com!timj
From: timj@tekchips.tek.com.UUCP
Newsgroups: mod.computers.apollo
Subject: Submission for mod-computers-apollo
Message-ID: <8703021625.AA02172@tekchips.TEK.COM>
Date: Mon, 2-Mar-87 11:25:35 EST
Article-I.D.: tekchips.8703021625.AA02172
Posted: Mon Mar  2 11:25:35 1987
Date-Received: Tue, 3-Mar-87 21:46:41 EST
Sender: daemon@ucbvax.BERKELEY.EDU
Distribution: world
Organization: The ARPA Internet
Lines: 52
Approved: apollo@yale-comix.arpa

Path: tekchips!timj
From: timj@tekchips.TEK.COM (Tim Johnson)
Newsgroups: mod.computers.apollo
Subject: Backup across TCP/IP
Keywords: backup
Message-ID: <1091@tekchips.TEK.COM>
Date: 2 Mar 87 16:25:34 GMT
Distribution: na
Organization: Tektronix Inc., Beaverton, Or.
Lines: 40

I do not advise using 'tar | rsh dd' for backing up files on the Apollos.  I 
was warned that characters can be lost.  Sure enough when I went back and
tried to read some of my backup tapes, tar was very unhappy with me.  The
occurance of these errors is a function of the load on the TCP/IP network
when the backup is being done.

I am presently VERY selectively backing up to cartridge tape ( I have 1.2 GIG
of mass storage ).

Apollo seems to be accepting the fact that in order for them to survive their
machines must operate in a non-homogeneous environment and are begining to
face up to problems such as backing-up across the TCP/IP network.

In our situation we have lots of UNIX and VMS systems with a several small
(3-5nodes)networks of Apollos.  There is no way we can rationalize spending
the bucks for a 9-track for three of four machines even though they typically
have .5 - 1.5 GIG of disk.  We do have two larger Apollo networks (>30 nodes)
that have their own 9-track.  All of these rings are physically and politically
separated and therefore cannot share 9-tracks.

The response I got from Apollo ranged from don't be so damn cheap (the old
Apollo folks speaking) to asking what our exesting environment is and what to
we need to solve the the problem (hopefully the new Apollo speaking).

By the way I came from a UNIX world and I find wbak a real joy to use and hope
Apollo enhances it to write to a socket.  A server could be started over on
the HOST the could read from this socket and preserve the blocking factor. A
person familar with sockets could write this tool in a day.  The hard part is
what to do about non BSD hosts that don't support sockets. I just hope Apollo
sees fit to release the easy part rather than wait for years to find the
`ideal solution'.  I suppose just a version of wbak that writes to standard out
would allow me to write the server on the host no matter what OS it was
running.

Maybe Apollo should just distribute sources for utilities such as wbak- but
then that is a whole other issue.


Tim Johnson
Tektronix
Beaverton, Oregon