myrinet problem

Hello everybody, I'm Cesar from Universidad Complutense de Madrid. I'm
sorry to send 
this mail to this distribution list, buy I don't know how to make the
myrinet works.
I contacted with [email protected] but I don't receive news of them, so
please, help me.

I'm trying to build a cluster, with Red-Hat 7.0, and four nodes. Each
node has 2 proccesors PemtiumIII 866MHZ.
 We want to use it to develop mpi
aplications over Gm,
and PBS. I don't know if the myrinet is working, I mean, I load the gm
driver in each node, 
(the green ligth of the switch turns on), and I load the mapper in
exactly one node. When I
see the mapper.map archive, all nodes are detected. 
I'm trying to ping to other host using myrinet and I receive "host
unreacheable"
(ping -I myri0 myri1)
When I run gm_anallize, it's seems works but when It stops, It doesn't
come back to the prompt and the port goes busy. I depured the network
with tcpdump, and
It's seems to be a problem of routing. The broadcast messages go by the
switch, but each node doesn't know
where are the others. I did not compile the linux kernel, It's the
default.

Thank you very much, and very sorry for this mail.


Hardware: boards PCI32
        switch mf2-sw8
	gm-driver: 1.2.3

uname -a:
Linux pila0 2.2.16-22enterprise #1 SMP Tue Aug 22 16:29:32 EDT 2000 i686
unknown                                                                   


My files are:

**********************+Ifconfig
eth0      Link encap:Ethernet  HWaddr 00:4F:4E:00:CC:6D
          inet addr:192.5.1.2  Bcast:192.5.1.255  Mask:255.255.255.0
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
          RX packets:935 errors:0 dropped:0 overruns:0 frame:0
          TX packets:799 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:100
          Interrupt:18 Base address:0xdc00
 
lo        Link encap:Local Loopback
          inet addr:127.0.0.1  Mask:255.0.0.0
          UP LOOPBACK RUNNING  MTU:3924  Metric:1
          RX packets:106 errors:0 dropped:0 overruns:0 frame:0
          TX packets:106 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:0
 
myri0     Link encap:Ethernet  HWaddr 00:60:DD:7F:F0:17
          inet addr:192.5.2.2  Bcast:192.5.2.255  Mask:255.255.255.0
          UP BROADCAST NOTRAILERS RUNNING MULTICAST  MTU:3752  Metric:1
          RX packets:0 errors:0 dropped:0 overruns:0 frame:0
          TX packets:0 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:100
          Interrupt:16  

*******************Route

Kernel IP routing table
Destination     Gateway         Genmask         Flags Metric Ref    Use
Iface
192.5.1.0       *               255.255.255.0   U     0      0        0
eth0
192.5.2.0       *               255.255.255.0   U     0      0        0
myri0
127.0.0.0       *               255.0.0.0       U     0      0        0
lo  


*******************/etc/hosts

127.0.0.1               localhost.localdomain localhost
192.5.1.1 castilla.dacya.ucm.es castilla
192.5.1.2 pila0
192.5.1.3 pila1
192.5.1.4 pila2
192.5.1.5 pila3
192.5.2.2 myri0
192.5.2.3 myri1
192.5.2.4 myri2
192.5.2.5 myri3              

******************/etc/hosts.equiv

castilla
localhost
castilla.dacya.ucm.es
pila0
pila1
pila2
pila3
myri0
myri1
myri2
myri3           



***************mapper.map
;23
 
h - "pila0"
1
0 s - "s0" 0
number 0
address 0060dd7ff017
gmId 1
hostType 0
 
h - "pila3"
1
0 s - "s0" 1
number 0
address 0060dd7fea17
gmId 3
hostType 0
 
h - "pila2"
1
0 s - "s0" 3
number 0
address 0060dd7fea07
gmId 3
hostType 0
 
h - "pila1"
1
0 s - "s0" 6
number 0
address 0060dd7fea11
gmId 4
hostType 0
 
s - "s0"
4
0 h - "pila0" 0
1 h - "pila3" 0
3 h - "pila2" 0
6 h - "pila1" 0
number 0               

**************pila0.host
1 0 0060dd7ff017 "pila0"
2 0 0060dd7fea17 "pila3"
3 0 0060dd7fea07 "pila2"
4 0 0060dd7fea11 "pila1"  



These are the outputs of the test tools.

cesar moreno iglesias escribió:
> 
> On Wed, 28 Mar 2001 [email protected] wrote:
> 
> >
> > Mr. Iglesias,
> >
> > When the mapper runs, does it report that it found the correct number of
> > hosts and switches for you network?
> Yes I think so, we have one switch and four hosts
> I run "mapper active.args " in pila0
> 
> found 4 hosts and 1 switches (5 nodes)
> sleeping for 30000 milliseconds
> 
> >
> > After running the mapper, run gm_board_info on each node.  Does each node
> > see every other node correctly?
> Yes, but the command hungs at the end. It doesn't comb back to the prompt
> 
> Out in pila0
> 
> Board number 0:
>   lanai_clockval = 0x90479047
>   lanai_cpu_version = 0x0403
>   lanai_board_id = 00:60:dd:7f:f0:17
>   lanai_sram_size = 0x00100000 (1024K bytes)
>   fpga_version = "Fri Mar 6 13:07:48 1998"
>   more_version = ""
>   board_type = 0x0001 (GM_MYRINET_BOARD_TYPE_1MEG_SRAM)
>   bus_type = 0x0002 (GM_MYRINET_BUS_PCI)
>   product_code = 0x0018
>   serial_number = 16607
>     (should be labeled: "M2F-PCI32c-16607") LANai time is 0x0287ee2e0
> ticks, or about 5 minutes since rese t. This is node 1 (pila0)
> node_type=0 Board has 8 ports and has space for 4171 nodes/routes. Port:
> Status PID
>    0:  BUSY 1152
>    1:  BUSY 1140
>    2:  BUSY 1157 (this process [gm_board_info])
>    3:  BUSY 1 (ip-over-GM)
>    4:  free
>    5:  free
>    6:  free
>    7:  free Route table for this node follows: The mapper 48-bit ID was:
> 00:60:dd:7f:f0:17 gmID MAC Address Hostname Route ---- -----------------
> -------------------------------- ---------------------
>    1 00:60:dd:7f:f0:17 pila0 80 (this node) (mapper)
>    2 00:60:dd:7f:ea:07 pila1 81
>    3 00:60:dd:7f:ea:11 pila2 82
>    4 00:60:dd:7f:ea:17 pila3 83            A
> <--------It hangs here
> 
> >
> > Try running "gm_allsize -g".  Does that complete?  If it doesn't complete,
> > hit Ctrl-C to interrupt it,
> 
> # ./gm_allsize -g
> 1 39.08 2 39.16 3 38.75 4 39.08 5 39.08 6 39.13 7 39.12
> 8 39.13 9 38.62 10 39.16 11 38.65 12 39.21 13 38.91 14 39.09 15 39.27 16
> 38.67 18 39.23 20 39.47 22 38.73 24 39.05 27 39.56 30 39.27 33 39.28 37
> 39.30 41 39.48 46 39.57 51 39.37 57 39.60 64 39.60 72 39.66 81 39.99 91
> 39.63 102 39.72 114 40.41 128 44.01 144 43.88 162 44.50 182 44.83 204
> 44.99 229 44.47 257 50.02 289 50.63 325 50.77 365 53.58 410 54.43 461
> 55.78 518 56.57 582 59.56 654 59.56 735 62.28 826 64.17 929 66.38 1045
> 70.34 1175 71.39 1321 75.74 1486 79.97 1671 87.29 1879 93.64 2113 102.09
> 2377 108.88 2674 118.61 3008 129.68 3384 140.94 3807 153.66 Total queued =
> 14600. Total sent = 14600.
> Total received = 14600.
> 
> -----It hangs here, and there is no response to control+c (neither kill
> -9)
> 
>  then run gm_debug and send us the output.
> >

Opening board 0, port 0

DMA rate for 4096 Byte transfers
        bus_read  (send) = 77 MBytes/s
        bus_write (recv) = 89 MBytes/s

Some counters from LANai
 
error counters
  drop_cnt                                            0
  no_match_for_recv_cnt                               0
  no_match_for_dgram_recv_cnt                         0
  no_match_for_ether_recv_cnt                         0
  no_match_for_raw_recv_cnt                           0
  out_of_sequence_cnt                                 0
  badcrc_cnt                                          0
  bad_header_cnt                                      0
  nack_receive_open_connection_cnt         0
  nack_receive_close_connection_cnt        0
    nack_ignore_close_connection_cnt         0
  nack_received_cnt                        0
  nack_normal_cnt                          0
  nack_ignored_cnt                         0
     (bad) packet_sexno 0000 0000
     (bad)    ack_sexno 0000 0000
 
  resend_cnt                               0
  nack_cnt                                 0
  nack_down_cnt                            0
  nack_reject_cnt                          0
  handle_connection_reset_request_cnt      0
    nack_send_open_connection_cnt            0
    nack_send_close_connection_cnt           0
    nack_send_nothing1_cnt                   0
    nack_send_nothing2_cnt                   0
 
 
global counters
  port[2].active_subport_cnt               0
                                                
<----It hangs here



-- 

Thank you very much.


********************************************************************
Cesar Moreno Iglesias
Universidad Complutense de Madrid
Departamento de Arquitectura de Computadores y Automatica
Para cualquier consulta de administracion 
e-mail: [email protected] ; [email protected]
Despacho 431 Fisicas

Tel: 91 394 70 71 
*******************************************************************