Uploaded image for project: 'WiredTiger'
  1. WiredTiger
  2. WT-10576

Return EBUSY on forced drop if there is an active transaction

XMLWordPrintableJSON

    • Storage Engines
    • 5
    • 2023-02-23 "Stoney Baloney", 2023-03-21 Ellen Ripley, 2023-04-04 Bibbidi-Bobbidi-Boo, 2023-05-30 - 7.0 Readiness, 2023-11-28 - Anthill Tiger, 2023-12-12 - Heisenbug, 2024-01-09 - I Grew Tired, StorEng - 2024-01-23, 2024-02-06 tapioooooooooooooca, 2024-02-20_A_near-death_puffin, 2024-03-05 - Claronald, 2024-03-19 - PacificOcean, 2024-04-02 - GreatMugshot, 나비 (nabi) - 2024-04-16, Nick - 2024-04-30

      Context
      The failure originates from BF-27630. The problem starts from the assert failure hit:

      WT_ASSERT(session, upd->txnid == txn->id || upd->txnid == WT_TXN_ABORTED);

      The assert gets hit when we are trying rollback an update on the update list within a transaction shown here:

                      if (S2C(session)->cache->hs_fileid != 0 &&
                  op->btree->id == S2C(session)->cache->hs_fileid)
                    break;
                WT_ASSERT(session, upd->txnid == txn->id || upd->txnid == WT_TXN_ABORTED);
                upd->txnid = WT_TXN_ABORTED;

      Looking at the BF and through GDB the upd variable shows as:

      (gdb) print *session->txn->mod->u.op_row.upd
$14 = WT_UPDATE:
  ('txnid', '14829735431805717965')
  ('durable_ts', '14829735431805717965')
  ('start_ts', '14829735431805717965')
  ('prev_durable_ts', '14829735431805717965')
  ('next', '0xcdcdcdcdcdcdcdcd')
  ('size', '3452816845')
  ('type', "205 '\\315'")
  ('prepare_state', "205 '\\315'")
  ('flags', "205 '\\315'")
  ('data', "0x7fa756be3bcf '\\315' <repeats 49 times>")

      The update shows as 0xcdcdcdcd, this means that the memory has been freed, which means when a transaction is accessing freed memory when trying to rollback an update from the update list. There has been further investigation under an ASAN build that the eviction module has a possible play here: 

      ] 0x6060001ac6c0 is located 0 bytes inside of 61-byte region [0x6060001ac6c0,0x6060001ac6fd)
[j0:n2] freed by thread T15 here:
[j0:n2]     #0 0x5559cbc8bc42 in free /data/mci/4c5523d6b930f0c1f82f5452d6add3b6/toolchain-builder/tmp/build-llvm-v4.sh-FAX/llvm-project-llvmorg/compiler-rt/lib/asan/asan_malloc_linux.cpp:127:3
[j0:n2]     #1 0x5559cfb41f4f in __wt_free_update_list /data/mci/fc587e29670e1d65f0277fab520f25c9/src/src/third_party/wiredtiger/src/btree/bt_discard.c:478:9
[j0:n2]     #2 0x5559cfb41f4f in __free_skip_list /data/mci/fc587e29670e1d65f0277fab520f25c9/src/src/third_party/wiredtiger/src/btree/bt_discard.c:438:13
[j0:n2]     #3 0x5559cfb41f4f in __free_skip_array /data/mci/fc587e29670e1d65f0277fab520f25c9/src/src/third_party/wiredtiger/src/btree/bt_discard.c:418:13
[j0:n2]     #4 0x5559cfb41f4f in __free_page_modify /data/mci/fc587e29670e1d65f0277fab520f25c9/src/src/third_party/wiredtiger/src/btree/bt_discard.c:222:13
[j0:n2]     #5 0x5559cfb41f4f in __wt_page_out /data/mci/fc587e29670e1d65f0277fab520f25c9/src/src/third_party/wiredtiger/src/btree/bt_discard.c:116:9
[j0:n2]     #6 0x5559cfc9e7ae in __wt_evict_file /data/mci/fc587e29670e1d65f0277fab520f25c9/src/src/third_party/wiredtiger/src/evict/evict_file.c:102:13
[j0:n2]     #7 0x5559cfa6d1a9 in __wt_conn_dhandle_close /data/mci/fc587e29670e1d65f0277fab520f25c9/src/src/third_party/wiredtiger/src/conn/conn_dhandle.c:434:9
[j0:n2]     #8 0x5559cfbd0202 in __sweep_discard_trees /data/mci/fc587e29670e1d65f0277fab520f25c9/src/src/third_party/wiredtiger/src/conn/conn_sweep.c:170:9
[j0:n2]     #9 0x5559cfbd0202 in __sweep_server /data/mci/fc587e29670e1d65f0277fab520f25c9/src/src/third_party/wiredtiger/src/conn/conn_sweep.c:400:9
[j0:n2]     #10 0x7f33a11cb2dd in start_thread (/lib64/libpthread.so.0+0x82dd)

            Assignee:
            tod.johnson@mongodb.com Tod Johnson
            Reporter:
            jie.chen@mongodb.com Jie Chen
            Votes:
            0 Vote for this issue
            Watchers:
            20 Start watching this issue

              Created:
              Updated:
              Resolved: