wiki:developers-meeting-20140910
Last modified 3 years ago Last modified on 09/10/14 16:46:37

[part of a series of meetings]

Participants

Gerd, Al, Paul, Tigran, Karsten

Agenda

[see box on the right-hand 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 :)

Karsten: Parallel reading of files from dCache

Al: Posted some patched, only few missing to commit the whole thing

Stress testing the alarms, looks satisfactory, but they are only meant for 2.11

Dmitry: Mostly opertional stuff, stuck transfers, globus online

Gerd: Executive meeting, writing board report, ...

Worked on migration module operational issues, debugging handshake timeouts debugging SRM semi-deadlock

Paul: GridKA School, dCache workshop + talk about federated identity

talked about future directions, preparing fixing releases

Tigran: JGlobus 2.1 + dCache, more complicated than expected

Problem with Voms-API, send fix to JGlobus, but no response, yet. Undid changes in dCache, adapted to new BouncyCastle? Had problems on dcache-cloud, ncount got out of sync,... investigation ongoing Got some patches from external contributors

Plans for patch-releases

Should we make a new patch release?

We want to get releases out, but currently some unit tests are failing Jenkins slaves died, etc.

Investigating.

Could this be because of increased concurrency and/or slowed down machines because of changed Virtual Machines state.

There is a patch to remove race-condition from tests.

But there are more.

  1. AtomicCounter? -> Gerd
  2. BasicTest? -> Tigran
  3. following Tests will be distributed via eMail

Trunk activity

Progress with new features...

questions to tigran

Bulk deletetion through SRM: Atlas showed some slides with 30Hz deletion rate from multiple clients. Gerd managed to archive single client performance of 188Hz

Atlas doing small file stages: Small Files scripts might help, but they are not ready to be distributed.

Globus Online

What's the status?

Will be able to run tests on a production system next week. Currently there is no culprit that causes the failures.

jGlobus

We need a new jGlobus release, but unit tests are failing.

Should we move away from jGlobus? We have some dependencies (principals), but other libraries provide similar functionality. Also srmClient depends.

Gerd: We could have different versions for what is awailable in the dists.

As soon as VOMS people upgrade, we get the new BouncyCastle? and also Canl.

Issues from [FIXME: Add link to yesterday's Tier-1 meeting]

Xavier, Marc, Gerd: Xavier discovered overloading pools. Advised to upgrade pool manager.

  • Billing file logs: Pool manager sending information to billing

Gerd: using 2.10.3 except for some problems

Outstanding RT Tickets

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

8442: This should rather be logged at INFO or DEBUG and change the output default pattern

The "suspended" case, there may be an error code missing

8427: Could this be caused by different root paths? -> get more config

=> We should improve the error message as well. => avoid logging it twice.

Review of RB requests

7183 -> Paul to have a look 7289 -> Karsten to have a look

Also please have a look at Al's Alarm patches

DTNM

Proposed: same time, next week.