Restore Socket after checkpoint

chercheur · 09-24-2012, 04:41 AM

Hello,

i have done the checkpoint of an application client server in C with BLCR (Berkeley Lab checkpoint restart), after a failure, i'd like to restart server (server.blcr) and client (client.blcr) but i should recreate sockets betwen new client and new server, have you an idea please ?

Thank you so much.

NevemTeve · 09-24-2012, 10:27 AM

Don't do any special; use socket(2), bind(2), connect(2), listen(2), etc. the usual way.

chercheur · 09-24-2012, 11:16 AM

How can i do that please ? because the server.blcr will restart on a new machine so haw the client.blcr will detect the new ip adress of the server

NevemTeve · 09-24-2012, 11:48 AM

I repeat: use the well-known socket functions to create a new TCP-connection. Of course it will not be the 'resurrection' of the old connection.

chercheur · 09-24-2012, 11:51 AM

yes i know haw to create socket , but the question haw can tell client.blcr about the new adress ip of the server? i think at first to developp a shared library .. what's your opinion ?

NevemTeve · 09-24-2012, 12:07 PM

I think restarting as such don't work together with network connection (nor shared files, databases etc) -- it is for long batch processes, to make them interruptible.

chercheur · 09-24-2012, 12:17 PM

the problème how the client.blcr will detect the new ip adress of the server? have you another idea please ?