-
Type: Bug
-
Resolution: Fixed
-
Priority: Major - P3
-
Affects Version/s: None
-
Component/s: None
Discovered while testing a workaround fix for WT-7833 and removing the flush_lock from tiered_abort. So this may need WT-7833 to be merged before this bug shows itself.
On verifying the directory after failure, tiered_abort looks for unexpected files left around after the forced crash. The point of failure looks like this. Sometimes the file name is "collection-00000000x.wtobj" rather than "shadow-....".
Program terminated with signal SIGABRT, Aborted. #0 0x00007fa68a26ef47 in raise () from /lib/x86_64-linux-gnu/libc.so.6 [Current thread is 1 (Thread 0x7fa68afa3340 (LWP 6530))] (gdb) up #4 0x00005614c53a9469 in verify_tiered (session=0x5614c70cedb0) at /home/ubuntu/wt/git/wt-7833-concurrent-tiered-writers/test/csuite/tiered_abort/main.c:610 610 testutil_assert( (gdb) list 605 * Logged tables, i.e. "oplog" or "local" may be unable to remove the last object 606 * from before the restart due to recovery applying log records. So if we get a stat 607 * return that indicates the file exists, verify it is one of those tables. 608 */ 609 if (i == last - 1 && ret == 0) 610 testutil_assert( 611 WT_PREFIX_MATCH(name, uri_local) || WT_PREFIX_MATCH(name, uri_oplog)); 612 else 613 testutil_assert(ret != 0); 614 /* Verify earlier objects exist in the bucket directory. */ (gdb) p name $1 = 0x5614c7f30d40 "shadow-0000000009.wtobj"