[openib-general] [PATCH] ib_sa_query tool

Hal Rosenstock halr at voltaire.com
Mon Mar 6 07:16:42 PST 2006


Hi Ira,

On Thu, 2006-03-02 at 18:23, Ira Weiny wrote:
> Hal, Makia, Trent and I spoke at the workshop about ways to resolve IB nodes in
> a cluster by name.  The included tool allows userspace to query the SA,
> optionally by "name" or node description.  The main push for us here at LLNL was
> to get the ability to do stuff like:
> 
> # odev7 /root > ibping `ib_sa_query -L odev3`
> Pong from odev3.(none) (Lid 0x18): time 0.193 ms
> Pong from odev3.(none) (Lid 0x18): time 0.110 ms
> 
> --- odev3.(none) (Lid 0x18) ibping statistics ---
> 2 packets transmitted, 2 received, 0% packet loss, time 1809 ms
> rtt min/avg/max = 0.110/0.151/0.193 ms
> 
> In essence use node names for the standard diag tools.  This uses the node
> description field which can be set at start up by a script for each node.  We
> realize this has some limitations but it seems useful here.

Thanks! This will be a helpful tool. 

Some comments and questions:

It appears to handle duplicated names but I haven't gotten far enough
with it to confirm this totally yet.

I don't understand the following in the break code
(break_get_all_records):

        for (i = 0; i < 5000; i++) {
                for (j = 0; j < 20; j += 50) {

It looks to me like the j loop will be done just once. Is that the
intention ?

Do you envision this tool to be expanded to cover other SA queries or
focused on just this ? (The tool name should be based on the answer to
this).

Also, I think this tool should be named saquery rather than ib_sa_query
(assuming it will be expanded to support more SA queries). OK ?

-- Hal




More information about the general mailing list