Percona XtraDB Cluster - HA scalable solution for MySQL

WSREP: conc slot cancel not supported

Reported by Raghavendra D Prabhu on 2013-10-02
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
MySQL patches by Codership
Undecided
Seppo Jaakola
Percona XtraDB Cluster
Status tracked in Trunk
5.6
Undecided
Unassigned
Trunk
Undecided
Unassigned

Bug Description

Getting this in error log a lot.

=======================================================

wsrep_srv_conc_cancel_wait(
/*==================*/
 trx_t* trx) /*!< in: transaction object associated with the
   thread */
{
#ifdef HAVE_ATOMIC_BUILTINS
 fprintf(stderr, "WSREP: conc slot cancel not supported\n");
#else
 os_fast_mutex_lock(&srv_conc_mutex);
 if (trx->wsrep_event) {
  if (wsrep_debug)
   fprintf(stderr, "WSREP: conc slot cancel\n");
  os_event_set(trx->wsrep_event);
 }
 os_fast_mutex_unlock(&srv_conc_mutex);
#endif
}

===============================================

this is in abort trx path.

Binaries on most of the modern platforms have atomic builtins enabled.

Changed in codership-mysql:
assignee: nobody → Seppo Jaakola (seppo-jaakola)

a) If this is redundant, then a

  if (wsrep_debug) { } is required since it is otherwise in the error log a lot.

b) srv_conc_enter_innodb_without_atomics seems to be only place where trx->wsrep_event is used, srv_conc_enter_innodb_without_atomics is called only when ATOMIC_BUILTINS are not defined. Hence, it makes sense to wrap wsrep_srv_conc_cancel_wait and its calls inside #ifdef HAVE_ATOMIC_BUILTINS as well.

Changed in percona-xtradb-cluster:
milestone: none → 5.6.13-24.0
milestone: 5.6.13-24.0 → none
Changed in codership-mysql:
status: New → In Progress
Seppo Jaakola (seppo-jaakola) wrote :

Moved high priority thread innodb concurrency check in higher level, high priority threads can enter with no limits.
Cancelling of threads waiting for innodb enter must be support both for with and without ATOMICS implementation.
Now WITH_ATOMICS case is sorted out so that BF aborting threads can enter innodb by force, after their os sleep ends.

Fix pushed in revision: http://bazaar.launchpad.net/~codership/codership-mysql/5.6/revision/3966

Changed in codership-mysql:
status: In Progress → Fix Committed
milestone: none → 5.6.14-24.1
Changed in codership-mysql:
status: Fix Committed → Fix Released
To post a comment you must log in.
This report contains Public information  Edit
Everyone can see this information.

Other bug subscribers