Uploaded image for project: 'Lustre'
  1. Lustre
  2. LU-5393

LBUG: (ost_handler.c:882:ost_brw_read()) ASSERTION( local_nb[i].rc == 0 ) failed

    XMLWordPrintable

Details

    • Bug
    • Resolution: Fixed
    • Major
    • Lustre 2.8.0
    • Lustre 2.4.3
    • RHEL 6 w/ patched kernel for Lustre
    • 3
    • 15009

    Description

      We are hitting this LBUG on one of our production systems recently updated to Lustre 2.4.3 (on 4th of June).

      LustreError: 11020:0:(ost_handler.c:882:ost_brw_read()) ASSERTION( local_nb[i].rc == 0 ) failed:
      LustreError: 11020:0:(ost_handler.c:882:ost_brw_read()) LBUG
      Pid: 11020, comm: ll_ost_io03_084
      
      Call Trace:
       [<ffffffffa07e6895>] libcfs_debug_dumpstack+0x55/0x80 [libcfs]
       [<ffffffffa07e6e97>] lbug_with_loc+0x47/0xb0 [libcfs]
       [<ffffffffa04a0d44>] ost_brw_read+0x12d4/0x1340 [ost]
       [<ffffffff81282c09>] ? cpumask_next_and+0x29/0x50
       [<ffffffff8105bf64>] ? find_busiest_group+0x244/0x9f0
       [<ffffffffa0abbf0c>] ? lustre_msg_get_version+0x8c/0x100 [ptlrpc]
       [<ffffffffa0abc068>] ? lustre_msg_check_version+0xe8/0x100 [ptlrpc]
       [<ffffffffa04a8038>] ost_handle+0x2ac8/0x48e0 [ost]
       [<ffffffffa0ac2c4b>] ? ptlrpc_update_export_timer+0x4b/0x560 [ptlrpc]
       [<ffffffffa0acb428>] ptlrpc_server_handle_request+0x398/0xc60 [ptlrpc]
       [<ffffffffa07e75de>] ? cfs_timer_arm+0xe/0x10 [libcfs]
       [<ffffffffa07f8d9f>] ? lc_watchdog_touch+0x6f/0x170 [libcfs]
       [<ffffffffa0ac2789>] ? ptlrpc_wait_event+0xa9/0x290 [ptlrpc]
       [<ffffffff81058bd3>] ? __wake_up+0x53/0x70
       [<ffffffffa0acc7be>] ptlrpc_main+0xace/0x1700 [ptlrpc]
       [<ffffffffa0acbcf0>] ? ptlrpc_main+0x0/0x1700 [ptlrpc]
       [<ffffffff8100c20a>] child_rip+0xa/0x20
       [<ffffffffa0acbcf0>] ? ptlrpc_main+0x0/0x1700 [ptlrpc]
       [<ffffffffa0acbcf0>] ? ptlrpc_main+0x0/0x1700 [ptlrpc]
       [<ffffffff8100c200>] ? child_rip+0x0/0x20
      
      Kernel panic - not syncing: LBUG
      

      The assertion is here:

       753 static int ost_brw_read(struct ptlrpc_request *req, struct obd_trans_info *oti)
       754 {
      ...
       879                 if (page_rc != local_nb[i].len) { /* short read */
       880                         /* All subsequent pages should be 0 */
       881                         while(++i < npages)
       882                                 LASSERT(local_nb[i].rc == 0);
       883                         break;
       884                 }
      

      I was able to get the content of local_nb from the crash dump.

      crash> struct ost_thread_local_cache 0xffff88105da00000
      struct ost_thread_local_cache {
        local = {{
            lnb_file_offset = 6010437632,
            lnb_page_offset = 0,
            len = 4096,
            flags = 0,
            page = 0xffffea0037309c10,
            dentry = 0xffff8808048c9a80,
            lnb_grant_used = 0,
            rc = 4096
          }, {
      ...
          }, {
            lnb_file_offset = 6010757120,
            lnb_page_offset = 0,
            len = 4096,
            flags = 0,
            page = 0xffffea00372ccf80,
            dentry = 0xffff8808048c9a80,
            lnb_grant_used = 0,
            rc = 512   <======== local_nb[i].rc != local_nb[i].len /* short read */
          }, {
            lnb_file_offset = 6010761216,
            lnb_page_offset = 0,
            len = 4096,
            flags = 0,
            page = 0xffffea0037176e98,
            dentry = 0xffff8808048c9a80,
            lnb_grant_used = 0,
            rc = 4096
          }, {
            lnb_file_offset = 1411710976,
            lnb_page_offset = 0,
            len = 4096,
            flags = 1120,
            page = 0x0,
            dentry = 0xffff8803690c2780,
            lnb_grant_used = 0,
            rc = 0
          }, {
            lnb_file_offset = 1411715072,
            lnb_page_offset = 0,
            len = 4096,
            flags = 1120,
            page = 0x0,
            dentry = 0xffff8803690c2780,
            lnb_grant_used = 0,
            rc = 0
          }, {
      ...
      

      This LBUG occurred 5 times since 06/06/14. Each time, we have a short read followed by some non-empty pages (rc != 0). You will find attached some output for the first 3 occurences.

      FYI, we have Lustre routers between the servers and clients as they are on a different LNET. Clients, routers and servers are running Lustre 2.4.3.

      Attachments

        1. 1407594033.bf26.debug.log
          8 kB
        2. 1407598675.bf27.debug.log
          23 kB
        3. 1407800513.bf33.debug.log
          7 kB
        4. MTS1351.tgz
          204 kB

        Issue Links

          Activity

            People

              bfaccini Bruno Faccini (Inactive)
              bruno.travouillon Bruno Travouillon (Inactive)
              Votes:
              0 Vote for this issue
              Watchers:
              10 Start watching this issue

              Dates

                Created:
                Updated:
                Resolved: