wiki:developers-meeting-20100317
Last modified 11 years ago Last modified on 03/17/10 17:26:15

[part of a series of meetings]

Participants

Timur, Vijay; Gerd; Owen, Jan, Paul, Tanja, Irina

Agenda

[see box to the side]

Postcards

Up to two minutes (uninterrupted) per person where they can answer two questions:

  • What I did last week (since the last meeting),
  • What I plan to do in the next week.

No questions until we get through everyone :)

Gerd: past-week: continued moving default from batch files to default files; upgrading NDGF to 1.9.7. next-week: 1.5 days at DESY next week.

Timur: perf. markers; confirm perf. markers are being sent even during the checksum calc. Best fix for problem in FTS. dcap vs pnfs. Next-week: outstanding RT tickets. Improper user authorisation, change to send whole auth. object to space-manager. After this making SRM survive database restart (fixed in other dCache code).

Vejay: 1) enstore collaboration. Want to test new Chimera integration. Writing doc. how to install and test. Timur asked to look into billing database, so looking at how to achieve this with web interface.

Jan: last week: integrated https into web interface -- managed to get certificate to gPlazma and got information back. Trying to get it all working with Spring. Tidying up code to production-quality, unit tests, etc.

Paul:

Owen: ETICS still and Hudson -- building automatic release process for dcap on many platforms.

Tanja: rewriting the pool to follow the new architecture. Working on how to introduce this with existing code-base.

Timur mentioned he'd written a non-blocking NIO stack for disk IO.

Pool when interacting with mover is still creating a thread. This is what we want to fix.

Irina: working on ACL test suite. Some tests in python for testing Unix permissions. Plan to extend these to include ACL tests. Updating ACL documentation. Working on tickets.

Status of work for 1.9.5

A (quick?) review of activity needed for the 1.9.5 release

srm-client:

Globus TCP port range supported? No, we have an option.

srm-client releases:

client patches, currently going into all branches, but we're releasing.

Currently release srm-client from 1.9.5.

Owen: shared SRM code as separate package? dcache-srm-client, dcache-srm-common, dcache-srm-server. The interface between dcache-srm-client and dcache-srm-common isn't stable.

Apply SRM client patches to 1.9.5 and trunk only (no other stable branches).

Right now:

dcap client code is in modules/dcap in dCache SVN, which also includes some Java code.

Plan to move the Java code into dCache server branch.

Status of work for 1.9.6

A (quick?) review of activity needed for the 1.9.6 release

Chimera cleaner patch from Gerd: 1.9.6-only.

Locking: new feature, so will appear in 1.9.8. Gerd: think this is locking on the namespace entry of the directory, but need to check.

Status of work for Trunk (a.k.a future 1.9.7)

A (quick?) review of activity needed for the 1.9.7 release

Upgraded to day found a number of minor issues. 1.9.7 good to release in next few days.

NDGF have been running stably now for 2 hours, with transfers.

Trunk

Gerd: continuing move declarations to defaults.

The parameter scans directory and loads them.

Duplicate property declarations should be detected and a warning emitted. Not tested though.

Review Board

Propose that if a patch has been waiting for a review for longer than one month then at the next meeting we assign someone who will review it. They should review it by the following meeting. We will check.

If the queue is shorter we can drop the time limit.

People can submitted the patch to stimulate discussion.

Some patches that are hard to review: controversial or too big.

Issues from yesterday's Tier-1 meeting

BNL problem

Gerd to investigate.

SRM restarts

This is likely a "feature" of the Server Socket Factory in Globus.

Moved to using a standard Globus,

Jetty, where we have our own connector, but probably inherits the bug from Globus as Gerd used the GLobus code as a template.

In GridFTPServer we read the cert. when the client connects: GSS negotiation happens after the client connects.

Makes sense to solve it generically for both Tomcat and Jetty.

Either Timur or Gerd to look into it. Whoever has time first sends the other an email.

Kopenhaven trip

Vijay: second invitation? Sent via UPS: scanned

OSG, talk by ES-Net that Verisign going to create infrastructure for allowing people to request a certificate

Head-count: 13

Use the train-trip to Germany to write a proposal of where we want

Then break this down into 1/2 person-day chunks of work. Also include a few 1 hour discussion.

Timur has created a

Main activity: solving the "user-management" problem.

Have a 1..1.5 hour slot at the end of the day with the discussion items. The rest of the time spend on the user-management problem.

Outstanding RT Tickets

[This is an auto-generated item. Don't add items here directly]

Review of RB requests

Paul will review 1589 after Jan updates the patch.

DTNM

Proposed: same time, next week.