[openib-general] Re: getting many (4/5 of the time) RDMA_CM_EVENT_ROUTE_ERROR

[Hal Rosenstock]

Hi Or,

On Wed, 2006-03-22 at 11:03, Or Gerlitz wrote:
> Hi Sean, Hal
> 
> Since updating to 2.6.16 and drivers/infiniband replaced by openib svn trunk 
> r5919 i am getting ETIMEDOUT (-110) cma event on the iser initiator side, 
> about 4/5 of my reconnect tries.
> 
> > iser:iser_cma_handler:event 0 conn ffff81003ee0be20 id ffff81007cd8e2f0
> > iser:iser_cma_handler:event 3 conn ffff81003ee0be20 id ffff81007cd8e2f0
> > iser:iser_cma_handler:event: 3, error: -110
> 
> Each time i get RDMA_CM_EVENT_ADDR_RESOLVED, so there's IB connectivity
> but then the path query failed.
> 
> I have not digged yet much into that, so i don't know if there's any
> issue at the SM/SA side (Voltaire SM/SA), will do that tomorrow.

Can you try it with OpenSM and see if the same thing happens ?

> Just curious, was there lately any change in along the cma/local-sa/rmpp/mad 
> layer chain that possibly might broke something? I recall there was a 
> change in the rmpp code, could it break something with the voltaire rmpp 
> impl?

Does this use RMPP ?

> i guess you have some synthetic test for RMPP, can i get it? (and 
> some info how to run it...)

There are 2 tests available:
1. In utils, there is grmpp
2. In osmtest, there is osmtest.

-- Hal

> Or.
>