CommitPolicy HARD can soak CPU
Bug #1036422 reported by
Peter Beaman
This bug affects 1 person
| Affects | Status | Importance | Assigned to | Milestone | |
|---|---|---|---|---|---|
| Akiban Persistit |
Fix Released
|
Critical
|
Peter Beaman | ||
Bug Description
While testing we observed high CPU utilization in JournalManager#
Related branches
lp:~pbeaman/akiban-persistit/fix-hard-commit-infinite-loop
- Nathan Williams: Approve
-
Diff: 194 lines (+58/-20)2 files modifiedsrc/main/java/com/persistit/JournalManager.java (+39/-20)
src/test/java/com/persistit/JournalManagerTest.java (+19/-0)
lp:~pbeaman/akiban-persistit/fix-hard-commit-infinite-loop-2
- Nathan Williams: Approve
-
Diff: 137 lines (+25/-15)4 files modifiedsrc/main/java/com/persistit/JournalManager.java (+7/-1)
src/main/java/com/persistit/Transaction.java (+14/-10)
src/test/java/com/persistit/JournalManagerTest.java (+1/-1)
src/test/java/com/persistit/stress/unit/CommitBench.java (+3/-3)
| Changed in akiban-persistit: | |
| assignee: | nobody → Peter Beaman (pbeaman) |
Revision history for this message
| Peter Beaman (pbeaman) wrote | #1 |
| Changed in akiban-persistit: | |
| status: | Confirmed → Fix Released |
| status: | Fix Released → In Progress |
| Changed in akiban-persistit: | |
| status: | In Progress → Fix Released |
To post a comment you must log in.
Revision 353 purported to fix this bug. However, it introduced two other problems:
(a) spinSleep waited longer (1ms) than the actual force() method required and therefore disk utilization was limited.
(b) in HARD mode, read-only transactions can wait forever for durability.