wiki:developers-meeting-20160127
Last modified 4 years ago Last modified on 01/27/16 16:35:53

[part of a series of meetings]

Participants

Al, Christian, Olufemi, Marina, Tigran, Paul, Jürgen, Dimitry

Agenda

[see box on the right-hand side]

Postcards

Al: Preparation of patches for resilience system, over 40 patches. Checking reason for DB scanning slowdown.

Christian: Human brain project and testing frameworks for dCache and web applications. FTS service is now monitored by Icinga.

Olufemi: Continued from last week.

Marina: Continued work on chunked uploads, web interface, Jenkins / release mgmt

Anupam: OpenID testing with Google as IDP

Tigran: Fixing NFS problems for DCAP. After meeting with Uni Hamburg: Kickoff document marks start of collaboration

Paul: Reviewed Ivans patches for the OwnCloud? sync client. BNL: Proxy problems. Contact with responsible persons at workshop. Wrote a patch for proxying remote transfers. Evaluation of ARC2's distributed ID management system and integration with dCache. Reworked the scripts controlling the release process in order to make it almost completely automated, removing all private information from the scripts. Continued work on QoS for Indigo Datacloud.

Jürgen: Bug fixes for SmallFiles?. Looked for problem with recognition of certs in latest JRE releases. Jenkins tests. Release management.

Dimitry: Did dCache upgrade, resulting in timeout messages from several components. After debuggin problem was resolved, no impact on trunk.

Special topics

Replica Manager

arossi: http://rt.dcache.org/Ticket/Display.html?id=8871

Ticket from BNL was about files that had their replica status disappear. It seems that there are problems in the structure of Replica Manager. There are doubts whether or not to consider this a data-loss-inducing issue. We discuss approaches and alternatives. For the moment, we will decide based on the experiences with fixing the issue for BNL.

Broken file:

Pools have no idea about client behaviour, and may not notice clients disappearing mid-transaction. This may lead to seemingly broken files from interrupted transfers. We'll add a message indicating that state to enable handling such cases. The door should retry in that case. The decision how to handle a case where all files are broken is still open.

Trunk activity

Paul reverted https://rb.dcache.org/r/8962/ since it broke several tests.

PIC had a problem with a stuck dCache instance. No heap dump or stack traces were done before restarting, so Paul and Tigran suggest creating those automatically on restarts to gather debug info.

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

Plans for patch-releases

Release was done today after fixing issues in the testing system.

Outstanding Documentation

Outstanding RT Tickets

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

Review of RB requests

New, noteworthy and other business

Jürgen: Managing Jenkins

Paul: 2.10 EOL coming up

DTNM

Proposed: same time, next week.