-
Type: Bug
-
Resolution: Fixed
-
Priority: Major - P3
-
Affects Version/s: None
-
Component/s: None
-
Storage Engines
Running schema_abort with tiered storage occasionally (1 in 10 runs on my virtual workstation) results in an apparent failure:
$ ./test_schema_abort -T 10 -t 5 -PT -Po dir_store Running test command: ./test_schema_abort -T 10 -t 5 -PT -Po dir_store Parent: compatibility: false, in-mem log sync: false, timestamp in use: true, tiered in use: true Parent: Create 10 threads; sleep 5 seconds CONFIG: test_schema_abort -PT -h WT_TEST.test_schema_abort -s 0 -T 10 -t 5 -PSD2053098,E9969491 Create checkpoint thread Create timestamp thread Create 10 writer threads Thread 0 starts at 0 Thread 5 starts at 5000000000 Thread 7 starts at 7000000000 Thread 3 starts at 3000000000 Thread 2 starts at 2000000000 Thread 1 starts at 1000000000 Thread 9 starts at 9000000000 Thread 4 starts at 4000000000 Thread 6 starts at 6000000000 Thread 8 starts at 8000000000 SET STABLE: 7c 124 Checkpoint 2084440 complete: Flush: NO. Minimum ts 124 Checkpoint 2084441 complete: Flush: NO. Minimum ts 2949 Checkpoint 2084442 complete: Flush: NO. Minimum ts 3479 Checkpoint 2084443 complete: Flush: YES. Minimum ts 5373 Finished a flush_tier Checkpoint 2084444 complete: Flush: NO. Minimum ts 6058 Checkpoint 2084445 complete: Flush: YES. Minimum ts 8546 [1702917105:433934][30535:0x7f6658fc2700], tiered-server: [WT_VERB_DEFAULT][ERROR]: void *__tiered_server(void *), 473: storage server error from tier_storage_copy: WT_NOTFOUND: item not found Finished a flush_tier [1702917105:433978][30535:0x7f6658fc2700], tiered-server: [WT_VERB_DEFAULT][ERROR]: void *__tiered_server(void *), 473: the process must exit and restart: WT_PANIC: WiredTiger library panic [1702917105:433994][30535:0x7f6658fc2700], tiered-server: [WT_VERB_DEFAULT][ERROR]: void __wt_abort(WT_SESSION_IMPL *), 28: aborting WiredTiger library Kill child Open database, run recovery and verify content Got stable_val 8596 812134 records verified $ echo $? 0
It turns out that despite the errors and apparent panic by the child process, schema_abort doesn't register this as an error. I.e., the test returns 0 rather than an error value.
This ticket is to investigate and fix the tiered storage failure.