Apparent Cache.refresh race condition #1211
Comments
|
Original comment posted by lowasser@google.com on 2012-12-07 at 11:32 PM (No comment entered for this change.) |
|
Original comment posted by j.e.frank on 2013-01-23 at 10:16 PM I ran into this same issue, with a cache that I am hitting hard at the time that a refreshAfterWrite is triggered. It ends up doing reload twice. |
|
Original comment posted by wasserman.louis on 2013-01-23 at 10:18 PM Are you hitting this issue with the 14.0-rc?? releases? This should have been fixed by 7508583 Status: |
|
Original comment posted by wasserman.louis on 2013-01-23 at 10:20 PM (No comment entered for this change.) Labels: |
|
Original comment posted by j.e.frank on 2013-01-23 at 10:29 PM Sorry, I didn't notice this issue was listed in the v14 release (I was just looking at its status here, which I see has now been changed to Fixed). No, I'm using 13.0.1. I'm evaluating Guava Cache as a replacement for Ehcache in a particular application. It's high profile (relatively speaking, for me) so I'm sticking with stable releases for now. |
|
Original comment posted by wasserman.louis on 2013-01-23 at 10:38 PM The 14.0 release should be happening relatively soon, FYI. |
gissuebot
added
type=defect
Original issue created by lowasser@google.com on 2012-11-23 at 01:37 AM
http://stackoverflow.com/q/13521523/869736
If I understand what's going on correctly, the issue is that in between the steps
can interleave badly with the end of a refresh:
A. Finish the refresh future.
B. Acquire the segment lock.
C. Set the value, using a StrongValueReference instead of a LoadingValueReference.
D. Update the write time.
E. Release the lock.
In particular, what can happen is that if a get happens just as the entry is done refreshing, the steps can interleave
A, B, 1, C, D, E, 2, 3, 4
in which case a refresh will get inappropriately scheduled: it will see that the write time was long enough ago that a refresh is appropriate, and then it will see that the entry has a StrongValueReference and is not currently loading, so it will start a refresh immediately after the previous refresh completed.
The text was updated successfully, but these errors were encountered: