[Open-Bioinformatics-Foundation] Notice: New OBF CVS server and disruptive infrastructure upgrades
Chris Dagdigian
dag@sonsorol.org
Tue, 1 Oct 2002 13:07:29 -0400
##
To avoid active OBF members getting many copies of this message we have
constructed an announce-only unique email list culled from all the
discussion lists that we operate. This email is going out to 1982
unique addresses. Questions about how/why you received this message
should be addressed to 'chris@bioteam.net'.
##
To all OBF affiliated developers and mailing list subscribers as well
as related parties like I3C.org, biopathways.org and emboss.org:
Subject: Significant server downtime beginning on October 2nd
Contents:
(1) Summary for non-developers
(2) Background info & history
(3) Details for developers and powerusers
OBF Developers, I3C.org members & emboss.org CVS users:
There is critical information that you must be aware of in the body of
this message.
Summary for non-developers
=======================
On Wednesday, October 2, 2002 (all day - EDT timezone) we are going to
relocate all of our servers including the primary web, email, FTP and
sourcecode repositories to a new (and faster) network. This move
requires changing all of our server IP addresses and physically moving
the machine to a building a short distance away from our current
colocation facilities located in Cambridge, Massachusetts. We are also
going to take advantage of this relocation to move our primary CVS
repositories on to a dedicated developer-access-only machine (details
below).
The server maintenance and physical relocation are expected to only
take a few hours. However it may take up to 48 hours for our new DNS
information to propagate around the world and replace the old
information that may be aggressively cached by institutions and ISPs.
For many of you the primary symptom will be a disruption in email
discussion lists that you may be participating in.
A good sign that the change has propagated to your network will is
being able to resolve the 'bioperl.org' domain name to something other
than 199.93.107.70
The reminder of this message is background material and specific
information for developers and account holders on our system.
We are asking in advance for forgiveness for any inconvenience this
rushed-to-avoid-disconnection transition will cause. There may be
issues beyond our control (Wyeth routers/firewall, reverse IP mappings
with upstream ISPs, domain names we don't control DNS for, etc.) that
could cause us minor problems over the next few days. Please hang in
there!
During the downtime/transition process Chris Dagdigian can be reached
directly at 1-617-877-5498 and 'chris@bioteam.net'. Feel free to
contact him at any time with questions, problem reports or requests for
status updates. Once email is flowing again you can also contact the
OBF sysadmin team at 'root-l@open-bio.org'.
Background & History
=================
One of our longest and most significant supporters has been Genetics
Institute (now integrated into Wyeth Research) -- for many years now
they have been allowing us to hang our servers off of their corporate
internet connection in Cambridge, MA.
Over the years this has been a non-trivial gift as the combined OBF
websites (particularly bioperl.org and biojava.org) consume a very
significant amount of internet bandwidth every day. More bandwidth than
we could probably afford to pay for, especially in the early days of
our project(s).
The connection we have been using for 4+ years is based on a redundant
pair of T1 lines connected to the Genuity/BBN backbone. This was the
primary corporate internet connection for Wyeth Research in Cambridge
until several months ago when the company brought in a new OC3 fiber
line offering 12mbit/second internet connectivity.
After the company moved its internet usage over to the OC3 line someone
at Wyeth began the process of canceling the T1 pair without the
knowledge that there were multiple outside-the-firewall servers still
dependent on that link.
Those of you curious as to why our servers fell off the internet for
several days a few weeks ago now know the answer as to what happened.
Our internet circuit was shut off by the carrier on orders from Wyeth
network admins who thought the lines were unused.
The disconnect caused an immediate panic within our sysadmin core -- we
knew what happened but were not at all sure if we were still welcome at
Wyeth Research anymore. We thought that we were going to be forced to
find a new home for our systems with zero notice.
Fortunately that was not the case.
Once Wyeth Research was alerted to our problem there were many
employees on both the scientific and IT side who interrupted their
busy schedules and went above and beyond their normal duties to (a)
bring the T1 circuit back temporarily and (b) work with senior IT staff
on a transition plan to bring our servers into an outside-the-firewall
managed DMZ zone on the new really fast OC3 internet link.
This is a partial list of people at Wyeth Research who helped us out:
Charles Richard III -- Vice President, Genomics
Steve Howes --- Director of Bioinformatics, Genomics
Steve Tenerelli --- Associate Director, Border Network Services
Rich Hollenbach -- Firewall Administrator, Border Network Services
Jason Blair -- IT Network Engineer
Rich DiNunno -- IT Network Engineer
Without the efforts of these and other folks we'd probably still be
frantically looking for a new hosting facility at this time.
Specifics for developers and account holders
===================================
(A) CRITICAL INFO FOR DEVELOPERS
We are going to take advantage of this planned downtime to bring a new
developer server online. You will not be able to CVS commit to the
current server after 11am EDT on October 2nd 2002.
The new CVS server will be called "dev.open-bio.org" and will be a Sun
Netra T1 server running Solaris 8 and directly attached to a large
external RAID storage array. As usual we will create DNS entries for
"dev.your-project-here.org".
DEV.OPEN-BIO.ORG will be for developers only and will be the new home
for ALL CVS SOURCECODE REPOSITORIES!
Anonymous CVS for non-developers will still be at 'cvs.open-bio.org'
and 'cvs.your-project-here.org'
The main reason for this move is to get our developers onto a platform
with hardware RAID protected storage. The additional benefit is that
the system will not be encumbered by any of our other web or email
services.
How the transition will be managed:
(1) The CVS repository on the current main server everyone uses will be
made read-only at 11am EDT on Wednesday October 2nd 2002.
(2) Shortly after we make the repository unwritable we will disable
inbound SSH access to the primary server so new connections or commits
will not be possible.
(3) The locked-down CVS repositories will be copied onto
dev.open-bio.org along with all user accounts, UIDs, GIDs, passwords
and home directories.
(4) The new machine will be moved to the new OC3 network and brought
online as 'dev.open-bio.org'. Additional DNS aliases will follow
(5) At the same time the CVS migration occurs we will be updating all
of out other servers and physically moving them to the new hosting
network.
The path of least resistance for developers is to:
(a) get your code changes committed prior to the lockdown deadline
(b) perform a clean checkout of the codebase from dev.open-bio.org when
it comes up and continue future work on it.
However it is highly likely that developers may have changes that they
will not be able or willing to commit prior to the lockdown deadline.
For those users the following example script has been suggested:
% find . -name "Root" ! -type d | xargs perl -p -i .orig -e
's/ext:(\S+):/ext:dev.open-bio.org:/
That unix shell command (untested) should put the proper
dev.open-bio.org entries into any existing CVS checkout directories
that you may have. If you are nervous about performing such an
operation then please wait until our servers are back online again and
query the OBF sysadmin mailing list ("rool-l@open-bio.org") to see if
someone can provide a script that has been used and tested by others.
For more information:
Since our email lists will be down if at any time you need more
information or updates on our progress please feel free to contact me
directly at 'chris@bioteam.net' or 1-617-877-5498,
(B) Domains that will be directly affected by our relocation
We control DNS for the following domains and will be in a position to
make quick changes when the IP transition occurs:
biodas.org
biofetch.org
biojava.org
biomoby.org
bioperl.org
biopython.org
bior.org
biosoap.org
biostandards.org
bioxml.org
open-bio.org
We do not control DNS for the following domains:
i3c.org -- They will need to update DNS information themselves or face
disruption
biopathways.org -- Not yet hosting on our servers so not a problem
emboss.org -- They only use us for CVS but need to be aware of the new
hostname for the cvs server