UpdateDQ2Jul5

Hi all,

Please stop your site service instances [with the account used to run
the services]:
dashb-agent-stop dq2agents
dashb-agent-stop dq2fetcher
dashb-agent-stop dq2udpserver

(in case agents don't stop properly you may kill all Python processes
- as the database will be scratched later anyway - and remove any
process lock files under /tmp/.s.dashboard.dq2*)

Then do [as root]:
apt-get update
apt-get upgrade

to your DQ2 site service installations.

After the code is upgraded, restart the database (since there has been
one schema change) [again with the account used to run the service]:
dq2site-recreate-database

Then restart the services [again with the account used to run the service]:
dashb-agent-start dq2udpserver
dashb-agent-start dq2fetcher
dashb-agent-start dq2agents

The upgrades solves the problem found when queries to the central
catalogues give back bad responses (the error is still logged, but it
is harmless).

Additionally, another problem looking up replicas for open datasets
has been fixed.

Let us know (atlas-dq2-dev@cern.ch) of any problems.

Known issues:
- subscriptions using ALL_SOURCES as the policy (which was mistakenly
used in Panda for some days) cause very slow lookups and multiple
errors as all ATLAS-wide catalogues are queried (some of which are not
accessible or were test instances);
- the database load is still excessive (particularly when doing
extensive file lookups) and next updates will address this problem.

This update has been tested on LCG sites as well as BNLPANDA.

cheers,
Miguel

-- RobertGardner - 05 Jul 2007