# Testing master to slave heartbeat protocol
# 
# Including:
# - user interface, grammar, checking the range and warnings about 
#   unreasonable values for the heartbeat period;
# - no rotation of relay log if heartbeat is less that slave_net_timeout
# - SHOW STATUS like 'Slave_received_heartbeats' action
# - SHOW STATUS like 'Slave_heartbeat_period' report

-- source include/have_log_bin.inc

connect (master,localhost,root,,test,$MASTER_MYPORT,$MASTER_MYSOCK);
connect (slave,localhost,root,,test,$SLAVE_MYPORT,$SLAVE_MYSOCK);

connection master;
reset master;

connection slave;

# there can be a warning if mtr set the global init value bigger than
# the following assignment. The warning is a grace for the user
# to warn her that the new value of the global var is bigger than
# the current value of the hb period, i.e the sanity condition
# 0 < hb < slave_net_timeout does not hold.
# The warning has to be disabled as mtr can change the init value
# for slave_net_timeout
--disable_warnings
--let SAVE_NET_TIMEOUT=`select @@global.slave_net_timeout`
set @@global.slave_net_timeout= 10;
--enable_warnings
###
### Checking the range
###

#
# default period slave_net_timeout/2
#
--replace_result $MASTER_MYPORT MASTER_PORT
eval change master to master_host='127.0.0.1',master_port=$MASTER_MYPORT, master_user='root';
--query_vertical show status like 'Slave_heartbeat_period';

#
# the max for the period is ULONG_MAX/1000; an attempt to exceed it is  denied
#
--replace_result $MASTER_MYPORT MASTER_PORT
--error ER_SLAVE_HEARTBEAT_VALUE_OUT_OF_RANGE
eval change master to master_host='127.0.0.1',master_port=$MASTER_MYPORT, master_user='root', master_heartbeat_period= 4294968;
--query_vertical show status like 'Slave_heartbeat_period';

#
# the min value for the period is 1 millisecond an attempt to assign a
# lesser will be warned with treating the value as zero
#
connection slave;
--replace_result $MASTER_MYPORT MASTER_PORT
eval change master to master_host='127.0.0.1',master_port=$MASTER_MYPORT, master_user='root', master_heartbeat_period= 0.0009999;
--query_vertical show status like 'Slave_heartbeat_period';

#
# the actual max and min must be accepted
#
--replace_result $MASTER_MYPORT MASTER_PORT
eval change master to master_host='127.0.0.1',master_port=$MASTER_MYPORT, master_user='root', master_heartbeat_period= 4294967;
--query_vertical show status like 'Slave_heartbeat_period';

--replace_result $MASTER_MYPORT MASTER_PORT
eval change master to master_host='127.0.0.1',master_port=$MASTER_MYPORT, master_user='root', master_heartbeat_period= 0.001;
--query_vertical show status like 'Slave_heartbeat_period';

reset slave;

#
# A warning if period greater than slave_net_timeout
#
set @@global.slave_net_timeout= 5;
--replace_result $MASTER_MYPORT MASTER_PORT
eval change master to master_host='127.0.0.1',master_port=$MASTER_MYPORT, master_user='root', master_heartbeat_period= 5.001;
--query_vertical show status like 'Slave_heartbeat_period';

reset slave;

#
# A warning if slave_net_timeout is set to less than the current HB period
#
set @@global.slave_net_timeout= 5;
--replace_result $MASTER_MYPORT MASTER_PORT
eval change master to master_host='127.0.0.1',master_port=$MASTER_MYPORT, master_user='root', master_heartbeat_period= 4;
--query_vertical show status like 'Slave_heartbeat_period';
set @@global.slave_net_timeout= 3 /* must be a warning */;

reset slave;


###
### checking no rotation
###

connection master;
--disable_warnings
drop table if exists t1;
--enable_warnings
#
# Even though master_heartbeat_period= 0.5 is 20 times less than
# @@global.slave_net_timeout= 10 in some circumstances master will
# not be able to send any heartbeat during the slave's net timeout
# and slave's relay log will rotate.
# The probability for such a scenario is pretty small so the following
# part is almost deterministic.
#

connection slave;
set @@global.slave_net_timeout= 10;
--replace_result $MASTER_MYPORT MASTER_PORT
# no error this time but rather a warning
eval change master to master_host='127.0.0.1',master_port=$MASTER_MYPORT, master_user='root', master_heartbeat_period= 0.5;
--query_vertical show status like 'Slave_heartbeat_period';

start slave;

connection master;
create table t1 (f1 int);

#connection slave;
sync_slave_with_master;
let $status_items= Master_Log_File, Relay_Master_Log_File;
source include/show_slave_status.inc;

# there is an explicit sleep lasting longer than slave_net_timeout
# to ensure that nothing will come to slave from master for that period.
# That would cause reconnecting and relaylog rotation w/o the fix i.e
# without a heartbeat received.

real_sleep 15;

# check (compare with the previous show's results) that no rotation happened
let $status_items= Master_Log_File, Relay_Master_Log_File;
source include/show_slave_status.inc;

###
### SHOW STATUS like 'Slave_heartbeat_period' and 'Slave_received_heartbeats'
###

--query_vertical show status like 'Slave_heartbeat_period';

#
# proof that there has been received at least one heartbeat;
# The exact number of received heartbeat is an indeterministic value
# and therefore it's not recorded into results.
#

let $slave_wait_param_counter= 300;
let $slave_value= query_get_value("SHOW STATUS like 'Slave_received_heartbeats'", Value, 1);
# Checking the fact that at least one heartbeat is received
while (`select $slave_value = 0`)
{
  dec $slave_wait_param_counter;
  if (!$slave_wait_param_counter)
  {
    --echo ERROR: failed while waiting for slave parameter $slave_param: $slave_param_value
    query_vertical show slave status;
    SHOW STATUS like 'Slave_received_heartbeats';
    exit;
  }
  sleep 0.1;
  let $slave_value= query_get_value("SHOW STATUS like 'Slave_received_heartbeats'", Value, 1);
}
--echo A heartbeat has been received by the slave
# cleanup

connection slave;
--disable_query_log
eval set @@global.slave_net_timeout= $SAVE_NET_TIMEOUT;
--enable_query_log

connection master;
drop table t1;

#connection slave;
sync_slave_with_master;

--source include/stop_slave.inc

--echo End of tests