Uploaded image for project: 'WiredTiger'
  1. WiredTiger
  2. WT-5356

WTPERF evict-btree-readonly crashes with a "Bad file descriptor" error

    • Storage Engines

      [This bug has been in develop as of October. I noticed it back then, but got distracted and forgot to report.]

      When we run the evict-btree wtperf workload, it crashes (almost immediately) with the following error:

      ----------------------------------------------

      ./bench/wtperf/wtperf -O ../bench/wtperf/runners/evict-btree-readonly.wtperf [1576882179:732026][17200:0x7fbb9b471700], statlog-server: __posix_file_write, 538: WT_TEST/./WiredTigerStat.20.14: handle-write: pwrite: failed to write 8201 bytes at offset 140186: Bad file descriptor[1576882179:732082][17200:0x7fbb9b471700], statlog-server: __statlog_server, 602: statistics log server error: Bad file descriptor[1576882179:732095][17200:0x7fbb9b471700], statlog-server: __wt_panic, 490: the process must exit and restart: WT_PANIC: WiredTiger library panicget_value in read.get_value in read. Error: WT_PANIC: WiredTiger library panic Error: WT_PANIC: WiredTiger library panicget_value in read. Error: WT_PANIC: WiredTiger library panicget_value in read. Error: WT_PANIC: WiredTiger library panic

      -------------------------------

      I turned on verbose and found that this happens quite predictably after the following sequence of events:

      • We open the WiredTigerStat file
      • We write to it a bunch of times
      • We close it
      • We open it for the second time
      • The very first time we try to write to it (after opening it for the second time), we get this error.

      I tested this on two different Ubuntu Linux systems (both 19.10) and got this error every time. I also got this error on OSX.

            Assignee:
            backlog-server-storage-engines [DO NOT USE] Backlog - Storage Engines Team
            Reporter:
            sasha.fedorova@mongodb.com Sasha Fedorova
            Votes:
            0 Vote for this issue
            Watchers:
            4 Start watching this issue

              Created:
              Updated:
              Resolved: