Should we create an RFE for that? 

 Here is why its important. We already have HA bits builtin that monitor 
processes for things like deadlocks or infinite loops. We also have signal 
handlers that inform the monitors that a core is going to be dumped so the 
monitoring gets turned off.

  But, there are some cases that cause cores without a signal being caught, 
like anything that calls thr_panic. What happens in those cases is that our 
monitor find the process to be non-responsive and since its hasnt been told a 
core is taking place it concludes the process is hung and kills it. Which 
disrupts the core dumping and leaves us with a corrupted core and thus unable 
to know what actually happened.

  So, having a supported means to determine that a core is being produced will 
allow us to keep our grubby hands off until its done. And figure out the root 
cause.

Regards
-d
 
 
--
This messages posted from opensolaris.org

Reply via email to