Skip to end of metadata
Go to start of metadata

You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 2 Next »

\uD83D\uDDD3 Date

\uD83D\uDC65 Participants

\uD83E\uDD45 Goals

  • Overview of open tickets

  • Consider new functionality / items

  • Detailed discussion of important topics

  • Site report activity

XRootD Kanban

key summary type created updated due assignee priority status resolution
Loading...
Refresh

\uD83D\uDDE3 Discussion topics

Item

Presenter

Notes

Async multi-io Buffering

James Walder

  • Implementation in python; some testing done; c++ ongoing.

Deletions

Ian Johnson

Investigating XRootD Slow Deletions Report from Ian. Repeated the tests from FTS studies.

  • Working on tests against a single host are ongoing. (in the FTS tests, unsure how the deletes are shared against the hosts in the dns alias).

  • To also consider extending the tests to the deletions of larger files than considered in the FTS tests (e.g. 5,10 – 100 GB).

  • Can the proxy and ceph daemon cause some additional latency ?


stat speed for xrootd/davs


Thomas, Jyothish (STFC,RAL,SC)


Report that 2k file stats were taking around 6hrs.

~ 1-1.5s for a stat via root vs 200ms for davs.

lockless reads.

Thomas, Jyothish (STFC,RAL,SC)

unlock on a stat call in libradosstriper, which didn’t have a lock, which might be the unlock seen in the metadata calls.

Tom’s observation that number of locks on the worker nodes has not been reduced.

Vector Reads, report from Rob Currie

All

Thomas, Jyothish (STFC,RAL,SC) Running tests with and without xrootd.async set on or off. (First tests might have failed due to grid proxy timeouts).

Did not discuss details this week. To digest and prepare sets of Actions / comments / tests from the outcome of the report.

Site reports

Lancaster

A very small amount of progress with escape token work for ska.


✅ Action items

  • James Walder Update Jira + RT; use of dev ceph ok for now, and low priority item for ceph team to move xrootd dev. to the res cluster.
  • To Confirm that the locks are removed appropriately from the WNs? Test WN to point at a dev cluster.
  • Discussion on Rob C. report postponed due to time.

⤴ Decisions

  • dev ceph instance is fine now for use as xrootd development instance
  • No labels