Details
-
Bug
-
Resolution: Fixed
-
Major - P3
-
None
-
None
-
Storage Engines 2018-09-10
Description
Running the new directio test that is in a pull request here: https://github.com/wiredtiger/wiredtiger/pull/4255. This test mixes a write workload with a lot of schema operations (create/rename/drop tables) within transactions. The test runner process periodically suspends the database writer process via SIGSTOP and using direct IO, copies the database home directory to a new location, roughly simulating a crash.
When recovering the copied directory, we see these errors:
Open database, run recovery and verify content
|
[1535549703:224768][21304:0x7f90be361740], file:A1308-4.wt, txn-recover: __posix_open_file, 711: WT_TEST.random-directio.CHECK/A1308-4.wt: handle-open: open: No such file or directory
|
[1535549703:224832][21304:0x7f90be361740], file:A1308-4.wt, txn-recover: __txn_op_apply, 278: operation apply failed during recovery: operation type 4 at LSN 1/3560320: No such file or directory
|
[1535549703:224861][21304:0x7f90be361740], file:A1308-4.wt, txn-recover: __wt_txn_recover, 721: Recovery failed: No such file or directory
|
test_random_directio: FAILED: check_db/619: wiredtiger_open(buf, NULL, ENV_CONFIG_REC, &conn): No such file or directory
|
This happens pretty regularly with this test, but is not completely predictable. I have a pre-copy of the failing directory.
Attachments
Issue Links
- duplicates
-
WT-2329 Crash during rename leaves metadata and file names out of sync
-
- Backlog
-