get_info environ caching is leaky and brittle
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
OpenStack Object Storage (swift) |
Fix Released
|
Medium
|
Unassigned |
Bug Description
Everytime we look at the get_info caching it seems like we should clean it up. We've even had the occasional bug slip through because it's hard to reason about what's going on in that sub-system because it crosses so many abstraction boundaries. Sometimes it just breaks down in surprising ways because when making unrelated changes.
Also it doesn't share cached data between subrequests.
A couple of good fixes popped up when trying to work through an unrelated change in lp bug #1449212 but they never got merged.
Some highlights:
* shallow-copy the WSGI environment
* do not rely on the proxy server setting the cache in the get_*_info() request environ, and instead use the response directly
... maybe some other good ideas, hopefully others can backfill all the stuff we wanted to fix in an itemized list for when we get around to trying to fix this on purpose.
Changed in swift: | |
importance: | Undecided → Medium |
Changed in swift: | |
status: | New → Confirmed |
Is there a way to reproduce the problems with get_info caching?