Ticket #1130 (closed clarification: duplicate)

Opened 13 months ago

Last modified 10 months ago

site freezes up during ingest and publish actions

Reported by: russ Owned by: rcave
Priority: high Milestone: 0.9.3_rc1
Component: ambra Version:
Keywords: Cc:

Description

compare 0.9.1 and 0.9

examine logs to try and explain differences

we suspect that performance of the overall site during ingest is worse in 0.9.1 than in 0.9

Change History

  Changed 12 months ago by russ

  • owner changed from russ to rich

  Changed 12 months ago by npeterson

  • owner changed from rich to dragisak

  Changed 12 months ago by npeterson

  • priority changed from high to low

  Changed 12 months ago by russ

  • owner changed from dragisak to russ

  Changed 11 months ago by russ

  • summary changed from test performance during ingest to site freezes up during ingest and publish actions
  • priority changed from low to high
  • version 0.9.1_rc1 deleted
  • milestone 0.9.2 deleted
  • owner changed from russ to npeterson
  • type changed from clarification to defect

reconfirmed for publish action in 0.9.2 - publish used to be REALLY quick before 0.9.1 - we're just flipping a bit and invalidating a few cache entries.

it's not a complete hang - once ingest/publish completes things work again. however, a lot of requests time out every time we publish/ingest

  Changed 11 months ago by russ

  • milestone set to 0.9.2

follow-up: ↓ 8   Changed 11 months ago by npeterson

  • owner changed from npeterson to ssterling
  • type changed from defect to clarification
  • milestone changed from 0.9.2 to 0.9.3

Pradeep has a fix to check in after 0.9.2 final. There is a deliverable for 0.9.3 for jmeter standardized performance testing. This work will be used to verify the fix.

in reply to: ↑ 7   Changed 11 months ago by pradeep

Replying to npeterson:

Pradeep has a fix to check in after 0.9.2 final.

The patch will only improve concurrency during lengthy txn.commit() calls - such as when Fedora API-M implementation has to copy blobs from the uploaded scratch area to the permanent storage. It will not impact anything else. ie. the txn serialization logic was added into 0.9. So if 0.9 release was fast and subsequent releases slow, there might be a problem elsewhere too.

  Changed 10 months ago by npeterson

  • owner changed from ssterling to rich

  Changed 10 months ago by rcave

  • status changed from new to closed
  • resolution set to duplicate

This is a duplicate of #1269

Note: See TracTickets for help on using tickets.