Skip to content
  • Shivji Kumar Jha's avatar
    340e0747
    BUG#14666177 - TIME DIFF BETWEEN LAST 2 RECONNECTS IS · 340e0747
    Shivji Kumar Jha authored
                   COMING > @@GLOBAL.NET_READ_TIMEOUT * 2 
    
    === Problem ===
    
    The test checks if the slave tries to reconnect before the
    value of slave_net_timeout. The check was done using the
    two variables:
    
    1) ts_last: timestamp of last reconnect-attempt by slave,
    2) ts_prev: timestamp of the second last reconnect-attempt
                by slave.
    
    The slave threads were killed using
         --source include/stop_slave.inc
    If the above statement is delayed due to CPU scheduling,
    the slave would get enough time to try two reconnects.
    If not, the slave threads would be killed after a single
    reconection attempt by the slave.In this case,
         ts_last=ts_prev.
    
    There was an assert condition given by:
         assert(@ts_last-@ts_prev >= slave_net_timeout
    where slave_net_timeout=10.
    
    In the above case,
         ts_last-ts_prev=0
    which is less than slave_net_timeout. So, the assert condition
    fired and the test failed.
     
    === Fix ===
    
    1) Deleted the line
         --source include/stop_slave.inc
       which killed the slave threads earlier.  
    2) The assert condition is replaced by a wait condition
       which makes sure that the slave makes two
       reconnection attempts.    
      
    340e0747
    BUG#14666177 - TIME DIFF BETWEEN LAST 2 RECONNECTS IS
    Shivji Kumar Jha authored
                   COMING > @@GLOBAL.NET_READ_TIMEOUT * 2 
    
    === Problem ===
    
    The test checks if the slave tries to reconnect before the
    value of slave_net_timeout. The check was done using the
    two variables:
    
    1) ts_last: timestamp of last reconnect-attempt by slave,
    2) ts_prev: timestamp of the second last reconnect-attempt
                by slave.
    
    The slave threads were killed using
         --source include/stop_slave.inc
    If the above statement is delayed due to CPU scheduling,
    the slave would get enough time to try two reconnects.
    If not, the slave threads would be killed after a single
    reconection attempt by the slave.In this case,
         ts_last=ts_prev.
    
    There was an assert condition given by:
         assert(@ts_last-@ts_prev >= slave_net_timeout
    where slave_net_timeout=10.
    
    In the above case,
         ts_last-ts_prev=0
    which is less than slave_net_timeout. So, the assert condition
    fired and the test failed.
     
    === Fix ===
    
    1) Deleted the line
         --source include/stop_slave.inc
       which killed the slave threads earlier.  
    2) The assert condition is replaced by a wait condition
       which makes sure that the slave makes two
       reconnection attempts.    
      
Loading