[okl4-developer] Question on thread/partition management

[Jean-Christophe DUBOIS]

 I guess my question might be a little out of scope because I don't
get that much reaction ...  

 So just curious: Is fault detection/management and fault
tolerance/recovery considered out of the scope of what OKL4 should
provide to the applications as services? Therefore it is left to the
user to develop/implement the way he sees fit ...   

 Regards   

 JC   

 On Thu, 24 Jul 2008 13:46:21 +0200, Jean-Christophe DUBOIS  wrote:  


 Hi Guys,  

 If a thread in one partition "crashes" is there any way to be
notified about this in the "root partition". For example If the
wombat/linux kernel has a kernel fault (divide by zero or anything
similar), is it possible to get notified about it in the "root
partition". How easy is it to determine which thread did crash?  

 I guess the above notification/exception mechanism would not cover
all cases. In particular it might happen that one thread would just
hang without necessarily raising an exception (or it would handle the
said exception
itself but would be in an unstable state after this). 
So an alternative (and maybe necessary) solution would be to
implément some kind of "heartbeat" between the various
partitions/servers/threads.That seems easy enough ...  

 The next question is, assuming you have determined which partition
crashed, how do you "restart" the thread(s) of this failing partition
from the root partition for example? Is there an easy way to get a
"clean" restart where all memory and environment is restored to the
initial state for this thread? More generaly is there a "service"
somewhere that would allow to load an ELF file dynamically in OKL4
(creating mempartion and all as required on the fly)?   

 Is this a supported scenario or are you better of restarting the all
thing (okl4 and everyting else)?  

 Thanks  

 JC   

   
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.okl4.org/pipermail/developer/attachments/20080729/73964057/attachment.htm