IC65531: DROPPING A DBSPACE WHILE A TABLE IS TRYING TO EXTEND ITS SPACE COULD CAUSE A DEADLOCK BETWEEN THE ONSPACE AND SQLEXEC THREAD

APAR status

• Closed as program error.

Error description

• The onspace thread holding the mutex chkpt, chunk table latch,
  is waiting for all threads
  to be out of a critical section.  While a sqlexec thread in
  critical section, mutex wait chkpt,
  waiting for the onspace thread to release the mutex chkpt,
  causing a deadlock and checkpoint hang.
  
  Using onstat -g lmx, you can see:
  1. thread 67, onspace holding mutex chkpt and in
  wait4critex()--waiting for all threads to exit critical section.
  
  2. thread 61, sqlexec in critical section (see the onstat -u and
  onstat -g ath | grep... below)
  waiting for mutex chkpt.
  
  Resulting in a deadlock between the two threads and hung
  checkpoint.
  
  
  $ onstat -g lmx
  
  IBM Informix Dynamic Server Version 11.50.FC6     -- On-Line --
  Up 02:50:12 -- 250384 Kbytes
  Blocked:DBS_DROP
  
  Locked mutexes:
  mid      addr             name               holder   lkcnt
  waiter   waittime
  9        70000000036ea28  dbs                67       0
  10       70000000036e9b0  chkpt              67       0      61
       29
  4865     7000000118f8028  pt_100145          61       0
  ...
  
  
  $ onstat -g stk 67
  
  IBM Informix Dynamic Server Version 11.50.FC6     -- On-Line --
  Up 02:50:32 -- 250384 Kbytes
  Blocked:DBS_DROP
  
  ...
  
  (oninit)yield_processor_mvp
  (oninit)mt_yield
  (oninit)wait4critex
  (oninit)isenter_critblock
  (oninit)rsdbsdrop
  (oninit)aud_isdbsdrop
  (oninit)dropspace
  (oninit)onspace
  (oninit)listen_verify
  (oninit)spawn_thread
  (oninit)startup
  
  $ onstat -g stk 61
  
  IBM Informix Dynamic Server Version 11.50.FC6     -- On-Line --
  Up 02:50:38 -- 250384 Kbytes
  Blocked:DBS_DROP
  
  ...
  
  (oninit)yield_processor_mvp
  (oninit)mt_lock_wait
  (oninit)mt_lock
  (oninit)challoc
  (oninit)dbsalloc
  (oninit)ptextend
  (oninit)ptbuild
  (oninit)rspnbuild_sz
  (oninit)sqispnbuild
  (oninit)bldpartn
  (oninit)creatab
  (oninit)aud_creatab
  (oninit)excommand
  (oninit)sq_execute
  (oninit)sqmain
  (oninit)listen_verify
  (oninit)spawn_thread
  (oninit)startup
  
  $ onstat -u
  
  IBM Informix Dynamic Server Version 11.50.FC6     -- On-Line --
  Up 02:50:42 -- 250384 Kbytes
  Blocked:DBS_DROP
  
  Userthreads
  address          flags   sessid   user     tty      wait
      tout locks nreads   nwrites
  7000000103bf028  ---P--D 1        informix -        0
      0    0     42       54
  ...
  7000000103c6338  S-BPX-- 30       informix 15
  70000000036e9b0  0    1     57       6
  ...
  
  $ onstat -g ath | grep 7000000103c6338
   61      70000001175d388  7000000103c6338  3    mutex wait chkpt
         3cpu         sqlexec
  
  This problem can be reproduced with a debugger.
  

Local fix

Problem summary

• ****************************************************************
  * USERS AFFECTED:                                              *
  * Users dropping a dbspace                                     *
  ****************************************************************
  * PROBLEM DESCRIPTION:                                         *
  * A dead lock can occur when  dropping a dbspace while a table *
  * is trying to allocate an extent in that dbspace.             *
  ****************************************************************
  * RECOMMENDATION:                                              *
  * Upgrade to IDS 11.50.UC7 and above.                          *
  ****************************************************************
  

Problem conclusion

• Problem first fixed in 11.50.xC7
  

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

  IBM IDS ENTRP E

• Fixed component ID

  5724L2304

Applicable component levels

• RB15 PSN

     UP