Posted by bdragon on January 28, 2008 at 11:11pm
| Project: | Drupal core |
| Version: | 5.x-dev |
| Component: | base system |
| Category: | bug report |
| Priority: | normal |
| Assigned: | Unassigned |
| Status: | closed (fixed) |
Issue Summary
If a site is busy at the time the garbage collector kicks in, it is possible for the system to have a meltdown if nothing finishes the garbage collect fast enough.
| Attachment | Size | Status | Test result | Operations |
|---|---|---|---|---|
| cache_meltdown.patch | 1023 bytes | Ignored: Check issue status. | None | None |
Comments
#1
Same for HEAD.
#2
1 line change, good to go, already live on drupal.org.
#3
This should go into D6 first.
#4
Great, committed to 6.x. RTBC for 5.x.
#5
Are you really sure this makes sense?
If a cache is not flushed successfully - the variable shouldn't tell it is flushed successfully. This change sounds very illogical and keeps stale data.
#6
Hass, the variable_set() never checked the success of the query, so there is no such change as you explained.
#7
Committed to 5.x.
#8
Automatically closed -- issue fixed for two weeks with no activity.
#9
Hi guys,
I think the "well rounded" / "good" behavior would be adding a global variable to determine "being flushed", and not use the "flushed" variable as it is confusing and erratic as hass correctly mentioned.
This is from a good practices point of view,
To your review,
Etamar
#10
etamar, doesn't work the fix for you?
#11
No answer, so a positive is assumed. Closed again.