[part of a series of meetings]

Participants

Gerd; Al, Dmitry, Gen; Christian, Antje, Jan, Paul, Tigranm Karsten, Tanja

Gerd: finished xrootd (mostly), started release v1.0, lots of tickets, started working on new job

Gen: nothing

Al: started working on gPlazma stuff; trying to understand gPlazma2 responsibility, a couple of points need clarifying ..

Dmitry: back to working on SRM requests. Instrumented SRM with liquibase which creates a simple table with indicies. Ran it all and ran it through a little bit stress-test. srmCopy is a bit of a problem. Unfortunately, the copy request, in addition to the state-machine, have lots of one-to-many maps. Now refactoring code to remove this constructs. srmPut and Get are OK.

Christian: package SRM Nagios probe and checked it was working. Put into ETICS, nightlies working, but mock-builds are failing (true for other packages). Looking at dependency issues. Problem: maven doesn't find dependencies when you execute certain goals.

Antje: still working on the SRM chapter [Dmitry please answer my questions!]

Jan: worked on Hudson, unit-tests are all there, emma is still causing problems. Pool-selection setup page in the webadmin, tripped up over trailing white-space.

Paul: hudson, support, RT, emails,

Tigran: NIS plugin for gPlazma2, Kerberos plugin for gPlazma2 . verified at DESY. This is for DESY Photon Science instance (no grid) but. Now working on ACLs.

VARARGS t_storageinfo table.. can we increase the length?

256 characters should be enough.

1.9.12 problem with the VARCHAR. Table created with CHAR(36) and script converts this to VARCHAR(36) but misses entries for ACL table. Tigran to fix this.

Test instance dCache + WebDAV as back-end for TSM backup utility.

Karsten: learnt some new corners in the dCache world.

Tanja: Tickets and writing documentation about WebDAV and kpwd file and finding out how it works. Updating patches. Worked on decorator for Chimera to check ACLs.

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 :)

Plans for patch-releases

Should we make a new patch release?

Shell script fix (Paul)

You need to have a keystore containing IGTF certificates. This is for setups that don't do client authentication: so no trust store is needed, but dCache doesn't start without one. WebDAV door probably has same problem.

Release 1.9.12, 2.0 and 1.9.5.

Gerd has a couple of patches in RB that he's like to be merged into supported branches.

Trunk activity

Progress with new features...

maven

Paul tries Maven 3 to see if this fixes the emma problem. If so, then deploy on Hudson.

In maven 2, running the dependency plugin directly doesn't work unless you also compile: something in the compile phase adjusts the dependency plugin so it works.

TSM

One of the funny behaviours of

They try to truncate the file (perhaps to detect renames).

PUT with zero length, then they do PUT with the contents.

One hack is to enable overwrite.

Isn't this just an option? Yes, but we always create a mover.

We don't do a redirect on PUT, so door sees what's happeninUg.

Another possibility is that they do a zero-length PUT to detect if you get a redirect.

Gerd thinks, if there, There's a specific HTTP header for doing the redirect on part: EXPECT

It could simply be a bug: the WebDAV library they're using does this.

Tigran will investigate further.

Try running a mac laptop. That does some very weird things over WebDAV. It issues some very command.

The TSM software provides a live synchronisation. This is for sync-ing e.g., laptops in real-time.

Update dcache

The dcache.conf updated to empty file. Can we get rid of it?

It's not completely empty: there's some comments at the top.

Can we remove this only for the Solaris build?

Do we need this file? Would we deliver a broken dCache if we don't have this file?

Issues from yesterday's Tier-1 meeting

Doris wants to find the content of the StorageInfo? dump

metaToYml (fails with pages of exceptions). Tigran to ping Gerd.

Gerd thinking of merging metaToYml support into dCache. Add command to pool to print storage-info stored on pool. As the file doesn't exist in the namespace, this is the only source of SI for this file.

LHCb

subset of space-token declared as "RELEASED". The space accounting updates linkgroup correctly. But, whenever they remove the file from the space (e.g., deleted file) the space is updated.

PIC seems.

SRM issue

At the moment, the only solution is to look

There was another manifestation: cleaner didn't remove the file,

What happens if you remove the file and, exactly at the moment the pool deletes the file's data, the message is lost. We don't retry sending it. For cleaner, we retry sending the delete message to pool, but space-manager is passive, so doesn't retry. This is probably not robust enough and should be fixed?

If we move the responsibility for sending update to space-manager from broadcaster to cleaner then it can retry if space-manager hasn't replied saying it has removed the file from its database.

Issues from EMI

Nothinig special

Outstanding RT Tickets

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

Dmitry: can you access GGUS ticket?

Review of RB requests

Suggestion for Al: shorter the patch, easier to get them in :-)

DTNM

Proposed: same time, next week.