wiki:developers-meeting-20091202
Last modified 11 years ago Last modified on 12/09/09 11:02:21

[part of a series of meetings]

Participants

Tigran, Jan, Paul, Irina; Gerd; Timur, Dmitry, Vijay

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 :) You can submit your postcard early, below:

Paul

  • Last week: worked on pnfsDump, gLite SE GLUE validation suite, info-provider related work
  • This week: more pnfsDump, info-provider (fix the *@* problem)

Gerd

  • Last week: Travel forms, finish WebDAV stuff (thanks Paul), tested new version of PureTLS, investigated how to solve the read PEP problem
  • Next week: Implement the solution for the read PEP problem

Timur

  • Last week: moved srm server code to log4j logging, solved problem with DDL.jdbc files creation in CWD. Made SRM startup automatically with terracotta. Investigated problems in in2p3.
  • Next week: RT Tickets, Terracotta and GSI performance testing. LQCD upgrade from 1.7.0 to 1.9.5. Preparation for CMS dCache upgrade.

Dmitry

  • Last week: a little work on the SRM ls, also linkgroupauthorisationfile (LGAF), matched using pattern-match. Tanya to test custom RPM. vogroup and vorole, loops over entries in LGAF. Problem could be that not all auth. files are being sent to space manager.
  • Next week: [...]

Vijay

  • Last week: couple of Timur patches; work on enstore, automatically detect underlying name-space provider.
  • This week: continuing work on enstore.

VOInfo and multiple roles (15+ Mins)

This may need a separate meeting to discuss how to proceed.

VO-Info message is sent:

Proposed: just forward whatever gPlazma sends to SpaceManager.

This information is present in the Subject. DN FQAN, mapped user and group .

To which entities the user was assigned.

Get a list of all possible.

GridFTP you might want to change directory, an operation that only one record is allowed to do then subsequent operation should use only that record (for example, space manager).

We have a chain of authorisation decisions; future operations should be based on those records that have succeeded until now.

Consider a user that is a member of two VOs: cms and atlas. CMS data is being written, but CMS is out of space. Do we use ATLAS space?

Maybe we should have this new call: "make implicit space reservation for me". Make calls to the SpaceManager calls explicit.

Tigran: I tried to fix multiple roles for gsi-dcap in 1.9.5.

Tigran: look to see how much needs changing. Try to send gPlazma auth information. Whatever I get as roles.

Imagine you have a CMS top-level path. Inside you have a directory that only ATLAS can write into. What is the correct semantics?

In Castor, these are called "Storage Info" and "File Family".

Space Manager ...

Physical space reserved for transfer.

Dedicate a meeting to discuss this further.

Primary group is used for ownership. Should this also be used for space reservation.

Current dcap reports single role, but multiple roles if configured in ...

Status of work for 1.9.5 (10 Mins)

What's needed before we do another release?

Some good improvements, so should release asap.

1.9.5-10 release date

Release tomorrow 11 am.

Status of work for 1.9.6 (15 Mins)

A (quick?) review of activity needed before we do the 1.9.6 cut.

Read permission checks fix

Need to update

Extend all PnfsMessages? with an ACE mask: this tells the PnfsManager what operations you want to do. So, if you want to read a file, then PnfsManager will check this for you.

Gerd and Timur to discuss how to update the SRM part of the logic.

SRM-ls

Once eliminated the

More progress by doing very detailed profiling of an individual tests.

Gerd's tests showed everything is faster with the new

Timur says Dmitry is now sufficiently happy about

For normal listing, this could be slower initially. For verbose listing, this should be much faster.

Continue discussion with Dmitry present in discussion.

Fix for the Chimera NFS start

Tigran: is it fixed now? Gerd: I'll check.

The error message looks like: "I support 2--4", so tries version 0, 1, 2, ...

The list clean-up

This is for WebDAV.

Handler inside Jetty to handle some part of name-space from a static storage area. Needs to be configurable.

Issues from yesterday's Tier-1 meeting (10 Min)

  • BNL reported an unexpected number of connections in internal interface of a door; six times the number.

Pedro sent some netstat information; Gerd asked for stack-trace information.

  • Pedro also asked what is the time-scale for fixing the pool migration/PinManager issue RT 5196?

PinManager? will (attempt to) remove the pin on the file. Should Something in a week's time.

  • Andrea asked about the IN2P3 SRM COPY issue. Is there anything we can do to speed up the investigation of this problem?

They have multiple JAR files in their directory. One of the RPM upgrades has failed. Asked them to move this directory and reinstall dCache. No reply yet from IN2P3. Both cog-url-1.6.0.jar, cog-url-1.7.0.jar and cog-url.jar.

RPM version is 1.9.5-6

Outstanding RT Tickets

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

RT 4721: Problem in GridftpClient? when copying files to dCache with srmcp

About using PowerPC not working with Cryptix library, with IceTea?: self-build requires the gcc/gcj.

Jan: not sure what we can do to help, but it would be rude to just close the hardware.

Tigran: I know that the University of Vancoover is running dCache on PowerPC.

Try to find out why it fails with IBM Java.

JGLobus people do minor changes in PureTLS, despite PureTLS project self-declared "dead" status.

Gerd: to forward new PureTLS jar file for Jan to check with his minimal test.

Tigran: to contact Uni. of Vancouver to ask which platform they use for dCache and how they do it.

RT 5286: Tar file

Problem is with ant. Only fix is to use the system tar.

Other

CSC reported ticket 5280

nlink problem in Chimera.

Globus gossip

Steve Tschki: Timur reported that they are going to be granted write access.

We should add Globus to review board.

Plan to switch from Apache Axis 1.0 and Apache CXF. (or CFX?)

Gerd: have tried jGlobus sockets in Jetty.

Review of RB requests

Delayed passive

GET / PUT commands could be solved by using the DELAYED PASSIVE command.

Should we implement "DELAYED PASSIVE"?

Gerd: why change just when sites and clients are starting to use GridFTP 2.0?

"DELAYED PASSIVE" isn't a standard.

DTNM

Same time, next week.