Details
-
Bug
-
Resolution: Fixed
-
Medium
-
None
-
eZ Publish 4.5, eZ DFS
Description
On a busy site, publishing a content object can lead to the viewcache being cleared and regenerated, with the stale content, before the full publish operation is completed.
Some Notes:
- The whole publishing operation is wrapped in a DB transaction.
- eZ DFS operations (cache clear) are immediate (separate transactions).
Steps to reproduce
Steps to reproduce, on eZ DFS:
- use apache benchmark to request an article on the public site - for example:
ab -c 2 -n 10000 http://localhost/public_site/link/to/article
- while the "stress test" is running, go to the admin UI and update/publish the article.
- stop the "stress test" and check the article on the public site
Result:
- The changes to the article won't be visible (instead, the previous version is cached).
- Different siteaccesses my display the updated content correctly, if cache hasn't been generated yet.
Attachments
Issue Links
- relates to
-
EZP-20558 Workflow Event/Approve not working correctly
-
- Closed
-
-
EZP-14088 Setting the version status to pending should be inside the db transaction used by the publish operation
-
- Closed
-
-
EZP-21563 workflows: adding approval to media tree causes fatal error
-
- Closed
-
-
EZP-21565 Workflow: restore object will cause fatal error
-
- Closed
-
-
EZP-20211 Clear view cache after post_publish operation
-
- Closed
-
-
EZP-20212 Add transaction steps to operations that need it
-
- Closed
-
- testing discovered
-
EZP-21599 Fix regression from EZP-19660 at a higher level
-
- Closed
-
- links to