You are here: Home User Information Facility Services Frontier Frontier Meetings for ATLAS Meeting Minutes Archive 2011 Minutes: 6/23/2011

Minutes: 6/23/2011

by John S. De Stefano Jr. last modified Jun 23, 2011 11:11 AM
Notes from the ATLAS Frontier meeting on June 23, 2011.

frontier-minutes-20110623.txt — Plain Text, 3 kB (3694 bytes)

File contents

Participants:   Dario Barberis, Florentin Bujor, John DeStefano, Alastair 
                Dewhurst, Dave Dykstra, David Front
Early check-in: Andrew Wong

*** Site status: ***

BNL:
- No issues
- Upgraded Frontier servlet to v3.29 a few weeks ago

CERN:
- Quattor template progress (Savannah #121536)
  * Dave, David, Sergey optimizing template sharing between experiments
- Log rotate problems with Frontier servlet, Tomcat packages
  * Waiting for new servlet RPM before upgrading production
    - Supports multiple servlets, improved configuration files, better 
      password handling in CDB (Quattor)

KIT:
- Frontier upgrade successful? servers report v3.29

LYON:
- N/A

RAL:
- Oxford Squid remains broken: was connecting directly to Frontier
  * Typo in squid.conf ACL for Oxford subnet
  * Discussion on fail-over Squid configuration recommendations:
    - Restrict site Squids to only the local site (CMS)
    - Allow restricted incoming ranges, with open destinations
    - Allow unrestricted incoming access, with destinations restricted to 
      local Frontier (recommended)
- Second Frontier server with latest servlet ready for deployment
  * Set up with VO access: can permit remote admin access given DN info 
  * Current server has 3.27
  * Recommend to deploy new server in place of existing, then upgrade existing 
    server before re-deploying it into production

TRIUMF:
- Upgraded our Frontier Servlet from 3.27 to 3.29
- No other issues

*** Deployment, Development, Monitoring, and Testing: ***

AGIS: 
- ADeS, ADiG working on examples to test client
- Still need API validation from Frontier perspective

Development:
- Upcoming updates to Frontier client
  * Currently the retry strategy only tries all servers through a proxy
    when there are server errors (which imply that the proxy is OK).
    Other errors including timeouts try all proxies with the first
    server, and then fail over to direct connections to each server.
    The strategy for timeout-like errors is now changing so all
    proxies will be tried with each server in turn before failing over
    to direct connections to each server.

Monitoring:
- SLS query errors from KIT, TRIUMF (403, no response, etc.)
  * Need to contact Frontier site admins regarding local ACLs
  * Dave advises sharing SLS polling script with CMS on frontier1
    instead of using Flavia's old scripts because it will be more
    reliable and stay up to date with any fixes that CMS puts in
    - SLS polls are executed from
	    /home/dbfrontier/local/periodic/every5m.job
    - CMS script is in
	    /home/dbfrontier/monit/slsfrontier/sls_frontier_new.pl
	and just needs to be updated to take 3 parameters:
	1. The names of the machines polled
	2. An option of whether or not to use smsget to see if 
	    machine is in maintenance mode (CERN only)
	3. The name of the xml output file which should go under
	   /home/dbfrontier/local/apache/frontier/sls/

Packages:
- Working on changes to servlet RPM
  * Support for multiple servlet instances (needed for CMS)
    - Single configuration file for multiple servlets
  * Optional, default variables: can be overridden by servlets
  * New Python configuration for servlets, handling passwords
    - Support for different Python versions
- Latest RPMs are ready for testing at non-CERN sites

Savannah:
- Dario closed 76089 (Frontier server overload), fixed in client v2.7.16

*** A.O.B.: ***

Savannah:
- No open bugs in Savannah for Frontier or Squid
- Shifters will be reminded to record bugs in Savannah
- Florentin should use Savannah to track progress

AS&C, July:
- May need talks for Frontier, CVMFS
  * Alastair will be at CERN and can present if needed
Document Actions
Filed under: , , ,