Subject: | |
From: | |
Reply To: | |
Date: | Tue, 7 Mar 2006 21:07:47 -0600 |
Content-Type: | TEXT/PLAIN |
Parts/Attachments: |
|
|
We need to have a downtime this week, and due to scheduling constraints,
it will be on Friday instead of our usual Thursday period.
Essentially all user access will be off for the entire day (6 AM - 6 PM),
however we will try to return individual parts back into service when they
are available. The UAF might be available by noon.
Here is the list of items we need to perform, some of which we have been
storing for some time. We did perform as many tasks as we could during the
recent unexpected IBRIX outages,
Downtime, Friday March 10
-------------------------
We need to upgrade our LCG installation to LCG 2.7 to be prepared for the
upcoming service and data challenges. This effectively disables all condor
worker nodes and LCG grid interfaces. This upgrade is significant and will
take a long time. As much as possible will be preloaded before the
downtime.
We need to install another network card in the main switch to increase our
throughput. This is a quick install, but networking needs to reboot the
main switch to ensure it works properly, and this means all connections
will be disrupted.
We need to swap out the infortrend disk arrays for the replacement nexsan
disk array for the physics space. This means IBRIX will be offline. After
we return to service, although all data will have been physically moved to
another disk array, logically it will still appear on /uscms_scratch. This
will take a few hours to complete.
We need to upgrade the firmware on many nexsan dCache disks arrays. We
need to upgrade the firmware on our old dCache disks units as well.
We need to add bonded GE connections to the faster/larger dCache pools.
We need to perform a postgres security updates on pnfs and dCache
databases.
dCache:
The dCache jars needs to be upgraded.
Hyperthreading on all dcache pools needs to be checked and turned off.
Space manager and billing manager setup files needs to be upgraded.
Breakeven needs to be changes to v5.
Upgrade postgres
Upgrade the billing and accounting configuration
---
USCMS Tier-1 Facility Downtimes will be announced on this mailing
list,[log in to unmask], and on the web at
http://www.uscms.org/SoftwareComputing/UserComputing/Downtimes/UAF_downtimes.html.
(The web version is updated whenever new information is available.)
During this construction and commissioning phase, many things need to be
updated regularly. We try to limit these updates to Thursday morning.
Generally, some part of the USCMS Tier-1 Facilities will be down every
Thursday morning. The team appreciates your patience. We hope to keep the
number and length of downtimes to the minimum possible set. The goal in
2006 in 1 major downtime per month.
Please send all questions and comments to [log in to unmask] or
[log in to unmask]
Jon
Jon A. Bakken [log in to unmask] (630) 840-4790
|
|
|