Return-Path: forrest
Received: by postgres.Berkeley.EDU (5.61/1.29)
	id AA12892; Fri, 7 Feb 92 13:45:04 -0800
Date: Fri, 7 Feb 92 13:45:04 -0800
Message-Id: <9202072145.AA12892@postgres.Berkeley.EDU>
From: forrest (Jon Forrest)
Subject: Important System Changes Coming Up
To: postgres@postgres.berkeley.edu
Sender: pg_adm@postgres.berkeley.edu

(This is being sent to you because you have an account on postgres or
hermes.)

There are several major system management events that will be taking
place in the next couple of months. By mentioning them now, I'm hoping
to reduce unpleasant surprises later. As you read this, keep in mind
that everything is subject to change. And, as always, suggestions and
complaints are welcome.

The first thing you should know about is that I'm going to be doing an
audit of all the accounts on postgres, hermes, and the workstations
owned by the Postgres group. The intent here is to make sure that the
people using resources on these machines are still contributing to
Postgres, Picasso, and/or Sequoia research. In a separate message I'll
be sending a small questionaire to all accounts that I don't
immediately recognize as belonging to active researchers. Mike
Stonebraker has agreed that this audit should be done. By the way, if
you received this message and you have no idea why, since you don't use
any of the postgres or picasso machines, please let me know and I'll
remove your account right away so that you won't be bothered by any more
messages like this.

Two large DECserver 5900 machines will be arriving sometime toward the
end of February. These machines were ordered as part of Project
Sequoia.  The task of serving home directories, currently performed by
hermes, will be taken over by one of these machines.  It should provide
substantially better performance than hermes and won't be taken down
unless there is an extremely important reason.  The other server will
be used to serve the new storage devices that we'll be acquiring and
will be the one that might be taken down more regularly.  So, one of
the first things you'll see is your home directory tree being
transfered to a new server. This will be done with warning so that any
machine that needs to mount the new filesystems can do so.  This step
should be painless.

After the home directories have been transfered, a pass will be made
over existing workstations to convert them into the new "Sequoia
Standard Ultrix" configuration. (Let me know if you want a copy of the
document that describes this.) This step has the most chance for
screwing up people's environments so it will be done carefully,
workstation by workstation.

In addition the the account audit, I'm going to go through aliases,
printcap entries, and the like to clean up the fluff that's been
accumulating over the years. I'll do my best to ask the experts before
deleting anything so, if everything goes right, you really shouldn't
notice any differences.  Nonstandard programs will be added on demand,
and won't necessarily be transfered by default. If you want something
let me know.

When hermes is no longer being relied on for NFS serving, it will be
upgraded to the latest version of Dynix. Right now we're about 4
releases behind, and we've also lost control of kernel configurations
so rebuilding a new kernel without causing something else to break is
difficult. This upgrade will actually be a complete installation of the
new system so the existing file system layout may change and existing
files will be deleted. Before this upgrade is done, any file systems
that will be affected this way will be copied to at least tape, and
maybe even to other file systems if disk space permits.

At some point after all the steps described above have been
accomplished, the Sun-3 that is now called postgres will be replaced by
a DecStation 3100, and the Sun-3 will be given away. I realize that the
postgres machine is used for several critical purposes, so it won't be
given away until after the DecStation incarnation of postgres is up and
running. In fact, the Sun-3 will stay up, using a different IP address
so if anything goes wrong we could fall back to the Sun-3.  One problem
with replacing the Sun-3 is that it currently is connected to one of
the last 9-track tape drives around. It is doubtful that we're going to
need to do much with 9-track tapes in the future, but in case we do,
arrangements will be made with the group that is going to take the
Sun-3 for us to have access to the tape drive. But, this access will
probably require special effort so be forewarned.

Each workstation will have a local disk that will contain a root, a
swap area, and a /var area. Any leftover space will be available for
whatever purpose the owner of the workstation wants to use it for.
These local disks will NOT be NFS-served so if there are things that
you think other users should have access to they should be put on a
disk connected to one of the servers.

Now a few words about backup policies. The home directory file systems
on the server will be backed up incrementally to tape every night.
File systems on local workstation disks will be backed up incrementally
to a disk on the server every night, and will then be backed up to tape
as part of the normal server backup procedure. This means that
restoring something from a local disk will be slightly more difficult
than restoring something from a server. This is a recognized
trade-off.

I'll try to make this transition as painless as possible but, given the
system management anarchy that currently exists, I'm sure there will be
rough spots. I'll try my best to let you know what they are ahead of
time. I hope you let me know about any special requirements you have
ahead of time so that you'll be a satisfied customer.

Jon Forrest 475 Evans 3-6764