[ofa-general] multi-rail subnet_prefix and MPI issue

[Frank Leers]

Hi,

I have a question regarding unique subnet prefixes for a 2-rail QDR  
configuration.  I'm using connectX on CentOS 5.3 with the distro- 
supplied OFED 1.3.2, kernel 2.6.18-128.el5, opensm-3.2.2-3.el5.

Any insight appreciated.

-frank




# ibstat
CA 'mlx4_0'
         CA type: MT26428
         Number of ports: 2
         Firmware version: 2.6.0
         Hardware version: a0
         Node GUID: 0x0003ba000100fc04
         System image GUID: 0x0003ba000100fc07
         Port 1:
                 State: Active
                 Physical state: LinkUp
                 Rate: 40
                 Base lid: 1
                 LMC: 0
                 SM lid: 1
                 Capability mask: 0x0251086a
                 Port GUID: 0x0003ba000100fc05
         Port 2:
                 State: Active
                 Physical state: LinkUp
                 Rate: 10
                 Base lid: 18
                 LMC: 0
                 SM lid: 18
                 Capability mask: 0x0251086a
                 Port GUID: 0x0003ba000100fc06

...and specify the unique subnet prefixes in each SM instance's config  
file like this:

# grep prefix /etc/ofed/opensm-ib0.conf
OPTIONS="$OPTIONS subnet_prefix 0xfe80000000000010"

# grep prefix /etc/ofed/opensm-ib1.conf
OPTIONS="$OPTIONS subnet_prefix 0xfe80000000000020"


Which results in these instances running:

# ps -ef|grep opensm
root      1498     1  0 05:16 ?        00:00:00 /usr/sbin/opensm -B - 
maxsmps 0 -f /var/log/opensm-ib0.log -p 15 -g 0x0003ba000100fc05  
subnet_prefix 0xfe80000000000010
root      2450     1  0 05:37 ?        00:00:00 /usr/sbin/opensm -B - 
maxsmps 0 -f /var/log/opensm-ib1.log -p 15 -g 0x0003ba000100fc06  
subnet_prefix 0xfe80000000000020


Yet we are seeing complaints from openMPI as such(single rail works  
fine):

$ mpirun -np 2 -machinefile ~/myduals ./IMB-MPI1
--------------------------------------------------------------------------
WARNING: There are more than one active ports on host 'n0051', but the
default subnet GID prefix was detected on more than one of these
ports.  If these ports are connected to different physical IB
networks, this configuration will fail in Open MPI.  This version of
Open MPI requires that every physically separate IB subnet that is
used between connected MPI processes must have different subnet ID
values.

Please see this FAQ entry for more details:

http://www.open-mpi.org/faq/?category=openfabrics#ofa-default-subnet-gid

NOTE: You can turn off this warning by setting the MCA parameter
     btl_openib_warn_default_gid_prefix to 0.
--------------------------------------------------------------------------
[[1372,1],0][btl_openib_component.c:2827:handle_wc] from n0051 to:  
n0052 error polling LP CQ with status RETRY EXCEEDED ERROR status  
number 12 for wr_id 197710080 opcode 0 qp_idx 0
--------------------------------------------------------------------------
The InfiniBand retry count between two MPI processes has been
exceeded.  "Retry count" is defined in the InfiniBand spec 1.2
(section 12.7.38):

   The total number of times that the sender wishes the receiver to
   retry timeout, packet sequence, etc. errors before posting a
   completion error.

This error typically means that there is something awry within the
InfiniBand fabric itself.  You should note the hosts on which this
error has occurred; it has been observed that rebooting or removing a
particular host from the job can sometimes resolve this issue.
Two MCA parameters can be used to control Open MPI's behavior with
respect to the retry count:

* btl_openib_ib_retry_count - The number of times the sender will
attempt to retry (defaulted to 7, the maximum value).
* btl_openib_ib_timeout - The local ACK timeout parameter (defaulted
to 10).  The actual timeout value used is calculated as:

    4.096 microseconds * (2^btl_openib_ib_timeout)

See the InfiniBand spec 1.2 (section 12.7.34) for more details.

Below is some information about the host that raised the error and the
peer to which it was connected:

Local host:   n0051
Local device: mlx4_0
Peer host:    n0052

You may need to consult with your system administrator to get this
problem fixed.
--------------------------------------------------------------------------
--------------------------------------------------------------------------
mpirun has exited due to process rank 0 with PID 9242 on
node n0051 exiting without calling "finalize". This may
have caused other processes in the application to be
terminated by signals sent by mpirun (as reported here).
--------------------------------------------------------------------------
[admin1:03239] 1 more process has sent help message help-mpi-btl- 
openib.txt / default subnet prefix
[admin1:03239] Set MCA parameter "orte_base_help_aggregate" to 0 to  
see all help / error messages