wiki:developers-meeting-20100310
Last modified 11 years ago Last modified on 03/11/10 16:11:12

[part of a series of meetings]

Participants

Gerd; Timur, Vijay, Gene; Paul, Owen, 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 :)

Owen: releasing to CERN SRM-client, dcap tagging of new release, ETIC, etic, etic... Setting up Hudson jobs.

Tanya: rewriting dcap door, investigating how pools are working, RE: webdav, Q: from Dima which client supports GET and PUT with https. Now looking into how to get certificates into it lftp client.

Irina: testing 1.9.6-4 and 1.9.7-1 concerning issue of some incorrect behavior of permission check (while copying files to dCache and reading files using gsidcap and webdav) that appears when permission check moved to PnfsManager (permissionPolicyEnforcementPoint=PnfsManager).

Timur: working on preparation and acting on for upgrade; OSG all-hands meeting. Reporting on storage meetings. Described work on Hadoop and Luster

Started looking at RT-ticket related to GridFTP performance markers.

Vijay: planning on working on enstore monitoring is working with chimera. encp is working, but en-move are failing. This is now fixed, but there are modules currently not including in the test-suite. Want to ensure monitoring is working correctly.

Dmitry: nothing to report.

Gerd: busy getting 1.9.7 ready. Been spending 1.5 days looking at memory usage regression in the pool.

Status of work for 1.9.5

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

Status of work for 1.9.6

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

Still needs a patch applied for the mounting. This is in review board.

Resubmitted it, previous.

Status of work for 1.9.7

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

YUM repository with 1.9.7 all the time.

The conversion script needs to include the kerberos- FTP and kerberos- dcap and weak- FTP doors.

Does the migration script support multiple doors?

The updates for services.sh is now only used by the migration script. This needs to be updated to support this.

Change FERMI.GOV to EXAMPLE.ORG in Kerberos.

Status of work for Trunk (a.k.a future 1.9.8)

A (quick?) review of activity needed for the 1.9.8 release

Biggest point: release date, 1st May.

Owen: I feel we should have longer releases between successive feature releases: 3 months would be better.

Gerd: support a 3-month release, but stick with 1st May for 1.9.8

Timur: 3-months would be good.

Tigran had stuff: schema migration, autoconf pool-manager, full proxy support from webdav.

Kerberos issue needs sorting.

Timur: preview releases, not supported but a snapshot of where we are going?

Gerd: there will be a rush to get certain "rough areas" fixed.

Billing/JSP web-dcache

What is the status of the billing monitoring servlet?

From this page, we say "Due to a possible security issue, the download of this tool has been disabled. We will provide an improved version soon." The links from this page are disabled, so no one can download the servlet.

The project appears to use content from the directory modules/dCache/diskCacheV111/scripts/plots/web-dcache/, but we have no infrastructure to build the sevlet from these directories.

The files in directory modules/dCache/di[...]/web-dcache/WEB-INF/classes/devel/ are broken as they have no package and are not in the root folder of the source directory (modules/dCache/di[...]/web-dcache/WEB-INF/classes/). Moreover, the classes do not appear to be referred to anywhere else. Are these classes "under development" code that shouldn't part of the production code-base?

Timur: this is the work of Vladimir, but feel this project is important. Don't know what the security issue: so don't know what needs to be fixed.

Would like to include this include this in standard.

Would also like to include SRM watch here.

Potential issue with database connection.

Billing could be used as debugging utility if we add query functionality.

Don't have information about complete life-cycle of files: store when file added to pool but not when a file is deleted from pool.

Gerd: should be easy to add; the info is present in the text file.

Another functionality CDF are asking for is when a file is requested, when file appears on the pool and when file is closed. Currently we store only when a file is closed.

Part of this information is already present in the info provided to the billing. None of the doors are filling in this information.

Use-cases:

historical behaviour, activity of current system.

This is something that needs implementing, but discussion this on the mailing list.

PNFS doesn't install on SL-5

Timur: yesterday, MIT college planning upgrade of their dCache and couldn't get the current RPMs to install RPM on Scientific Linux 5.3 due to missing libraries. On Monday he'll submit an RT ticket; need to release for eac.

Wait on RT ticket.

Java 1.6 and Client

Owen: have compiled the Java client with Java v6, but this made v6 byte-code. Are we saying that we only support Java v6?

Owen: I'll try to get it through as Java 1.6.

Timur: lets do 1.6 and see who complains.

Form and purpose of these meetings

Gerd: continued frustration about the length of the meetings. I feel these meetings are for: reporting progress, deligating deciding policies

Reporting bugs: we have ..

Extensive technical discussions: mailing

People should be responsible for investigating

Owen disagrees ... [...]

Timur: we have some issue with time limits; focus on issue that cannot resolved locally. The marking of tickets

For status of the different branches,

Do we know which open tickets are for 1.9.5 branch?

Owen: want to know when we have a release.

Have a release manager for 1.9.5, so we know when a release is due.

Information is available: we need someone who recommends when we should do a release. Have a vote.

Start a trac page with template-agenda with description; look at what is already.

ACL test suite

deferred to next week.

Performance markers problem?

Timur: looked into how performance markers work. It's a little tricky to understand as we have only one performance. Timer started in FTP door, executes some code that sends the performance markers to the client. There are conditions that must be match before the marker is sent; e.g., if move is not running then no performance marker is sent. PoolV4, checksum is calculated by pool, not mover and the mover has died.

Gerd: are you saying, during checksum calculation, "mover ls -binary" doesn't output anything at that point? I thought we did continued sending perf. markers while calculating checksums.

The problem is with compatibility with FTS. This is new behaviour in v2.1. Lets discuss this on the team list.

Timur: if you use standard gridftp client then it will print perf. markers on stdout.

Timur will investigate that we send performance markers.

Issues from yesterday's Tier-1 meeting

None: the meeting was very quiet.

Outstanding RT Tickets

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

RT 5461: pool startup problem - out of memory

This ticket is not about this problem. It is a feature request to support more robust startup.

Jon said these were generated by file-system corruption.

Gerd: promised to make code more robust.

RT 5489: kerberos-dcap, gplazma

Tigran will work on kerberos dcap door to follow how kerberos FTP door does it.

Don't discuss this ticket any more.

RT 5502: Peculiar problem

Ticket to be closed.

RT 5544: modification time of the file on chimera/restore from tape

Assign to Tigran or Irina.

Drop DEV-discuss.

RT 5549: change in the format for "ls" in srm cell/release notes

Complaint that we didn't documented the change.

Gerd to update the release notes.

If a code commit changes the format of the admin interface it should be mentioned in the SVN commit message.

Review of RB requests

DTNM

Proposed: same time, next week.