[Freeipmi-devel] Re: [llnl-devel] FreeIPMI Session Management

[Albert Chu]

Hey AB,

This implementation is something I hadn't really thought much about. 
But after reading the beginning of chapter 12 in the IPMI specification,
I think its something that could work and would be a lot easier to
implement.  I need to think about it more, but thanks for bringing it up.  

Al

----- Original Message -----
From: Anand Babu <address@hidden>
Date: Friday, November 14, 2003 8:01 am
Subject: [llnl-devel] FreeIPMI Session Management

> 
> Keeping the session open all the time:
> -------------------------------------
> Why we needed this?
> A way to determine if the BMC is still accessible via IPMI.
> 
> Not the right approach. Session defined by IPMI is a logical
> connection. Actual transport still happens over rmcp/udp/ip in
> datagram mode.
> 
> Even if you keep the session open, you will still have to send packets
> in connection less mode. When you don't get response you get to know
> that system is knocked out.
> 
> IPMI session requires every ~< 1 min keep alive call.
> Keep alive call is nothing but re-activating the session using same
> activate session call. 
> 
> Also session is a costly resource for BMC. Only 4 sessions are
> allowed. Implementor may choose to support only 1 session at a given
> time and still be compliant with IPMI spec.
> 
> Thats why I did not implement session_keep_alive(...) higher level
> call on top of activate session. Instead a better approach is
> to use the ipmi_ping call (higher level call in ipmi-wrapper.c)
> to determine if the system is still accessable.
> You can call ipmi_ping at regular cycles and keep an active list of
> systems in a list.
> 
> And sessions are short-lived. 
>  open_session (...)
>  chassis_ctrl (...)
>  close_session (...)
> (Note: you can find these higher level calls in ipmi-wrapper.c)
> 
> Al, Jim comments?
> 
> -- 
> Anand Babu
> CaliforniaDigital.com
> Office# +1-510-687-7045
> Cell# +1-510-396-0717
> Home# +1-510-894-0586
> 
> Free as in Freedom <www.gnu.org>
> 
> _______________________________________________
> llnl-devel mailing list
> address@hidden
> http://californiadigital.com/cgi-bin/mailman/listinfo/llnl-devel
>