IC78576: DEADLOCK CAN OCCUR BETWEEN BTSCANNER AND SQLEXEC THREAD THAT IS ROLLING BACK

APAR status

• Closed as program error.

Error description

• This problem was encountered on 11.50.FC4 and 11.70.FC2.
  
  
  When the deadlock occurs, your btscanner thread will have a
  stack similar to:
  
  IBM Informix Dynamic Server Version 11.70.FC2 -- On-Line
  (LONGTX) -- Up 00:33:43 -- 244644 Kbytes
  Blocked:LONGTX
  
  Stack for thread: 118 btscanner_0
   base: 0x000000004d61e000
    len:   69632
     pc: 0x0000000001216e5a
    tos: 0x000000004d62ea00
  state: cond wait
     vp: 4
  
  (oninit) yield_processor_mvp
  (oninit) mt_fast_wait
  (oninit) critwait
  (oninit) isenter
  (oninit) btc_range_scan_clean_buffer
  (oninit) btc_scan_index
  (oninit) btscanner_loop
  (oninit) startup
  
  You may or may not see the function btc_alice_scan_index in the
  above stack being called from btc_scan_index.
  
  
  and the rolling back session will have a stack similar to:
  
  IBM Informix Dynamic Server Version 11.70.FC2 -- On-Line (CKPT
  REQ) (LONGTX) -- Up 01:31:52 -- 244644 Kbytes
  Blocked:CKPT LONGTX
  
  Stack for thread: 79 sqlexec
   base: 0x000000004d891000
    len:   69632
     pc: 0x0000000001216e5a
    tos: 0x000000004d8a01e0
  state: cond wait
     vp: 3
  
  (oninit) yield_processor_mvp
  (oninit) mt_wait
  (oninit) btc_disallow_cleaning
  (oninit) rspnlockid
  (oninit) logundo
  (oninit) rlogm_undo
  (oninit) rollback
  (oninit) rsrollback
  (oninit) sqisrollback
  (oninit) sqrollback
  (oninit) net_buf_get
  (oninit) recvsocket
  (oninit) slSQIrecv
  (oninit) pfRecv
  (oninit) asfRecv
  (oninit) ASF_Call
  (oninit) _iread
  (oninit) _igetint
  (oninit) sqmain
  (oninit) spawn_thread
  (oninit) startup
  
  
  Eventually, IDS will hang at a CKPT REQ with onstat -g con
  showing the conditions: block, btc_partp and cp:
  
  $ onstat -g con
  
  IBM Informix Dynamic Server Version 11.70.FC2 -- On-Line (CKPT
  REQ) (LONGTX) -- Up 01:53:53 -- 244644 Kbytes
  Blocked:CKPT LONGTX
  
  Conditions with waiters:
  cid      addr             name               waiter   waittime
  5        44249f00         block              118      5052
                                       22       5018
  826      44249d88         cp                 121      9
  839      4c14dde8         btc_partp          79       4923
  1134     4424b990         bp_cond            30       5134
  

Local fix

Problem summary

• ****************************************************************
  * USERS AFFECTED:                                              *
  * All Users                                                    *
  ****************************************************************
  * PROBLEM DESCRIPTION:                                         *
  * Informix server can hang due to dead lock between sqlexec    *
  * thread that is rolling back a long transaction and btscanner *
  * thread that is in the middle of cleaning deleted items from  *
  * index pages.                                                 *
  ****************************************************************
  * RECOMMENDATION:                                              *
  * Please upgrade to 11.50.xC10 or higher when possible.        *
  ****************************************************************
  

Problem conclusion

• The problem is first fixed in IDS 11.50.xC10.
  

Temporary fix

Comments

APAR Information

• APAR is sysrouted FROM one or more of the following:

• APAR is sysrouted TO one or more of the following:

Fix information

• Fixed component name

  INFORMIX SERVER

• Fixed component ID

  5725A3900

Applicable component levels

• RB50 PSY

     UP