Network Testing and Debugging for AGLT2

During the last year we have seen many indications that all is not right with our network connections to BNL (and perhaps other sites). The most graphic example of the problem was observed during preparations for SC08 Terapaths demos. We had Hiro (BNL) start 10 "production" transfers from AGLT2 to NET2. Each of the 10 transfers was running at around 8 Mbits/second. A Terapaths reservation was put in place, reserving 1 Gigabit between AGLT2 and NET2. As soon as the reservation was activated the transfers increased from 8 Mbits/second to 12MBytes/second (total of 120 Mbytes/sec for 10 flows ~ 1 Gbit/sec). Since the end systems/applications running were the same before and after the reservation we can conclude that it was the change in network path corresponding to the reservation that made the difference.

Recently poor transfer performance moving files from AGLT2 to BNL has prompted us to try to debug the issue and hopefully isolate it so it can be fixed.

We have the following resources to use:

  • Two hosts at BNL, both with 10GE NICs: nettest10g.usatlas.bnl.gov and newmon.bnl.gov
  • One host at StarLight/Chicago with a 10GE NIC: wood1-chi.uslhcnet.org (10GE at 192.84.86.10)
  • Multiple 10GE hosts at AGLT2:
    • UM: umfs11.aglt2.org, umfs12.aglt2.org umfs13.aglt2.org
    • MSU: msufs05.aglt2.org msufs08.aglt2.org

In addition we have perfSONAR boxes at BNL, UM(psum01.aglt2.org, psum02.aglt2.org) and MSU (psmsu01.aglt2.org, psmsu02.aglt2.org)

Initial Info

I needed to setup explicit routes for wood1-chi.uslhcnet.org to insure the 10GE NIC was used to test to/from the subnets of interest. I added:
  • route add -net 192.41.230.0/23 eth2 ! (AGLT2)
  • route add -net 130.199.0.0/16 eth2 ! (BNL)
  • route add -net 192.12.15.0/24 eth2 ! (BNL/usatlas)

[root@wood1-chi ~]# route
Kernel IP routing table
Destination     Gateway         Genmask         Flags Metric Ref    Use Iface
192.65.196.64   *               255.255.255.224 U     0      0        0 eth0
192.84.86.0     *               255.255.255.224 U     0      0        0 eth2
192.12.15.0     *               255.255.255.0   U     0      0        0 eth2
192.41.230.0    *               255.255.254.0   U     0      0        0 eth2
130.199.0.0     *               255.255.0.0     U     0      0        0 eth2
169.254.0.0     *               255.255.0.0     U     0      0        0 eth2
default         E600chi-prod.us 0.0.0.0         UG    0      0        0 eth0

The current traceroutes between machines are shown here.

root@umfs11 ~# traceroute 192.84.86.10
traceroute to 192.84.86.10 (192.84.86.10), 30 hops max, 46 byte packets
 1  vl4001-nile.aglt2.or.230.41.192.in-addr.arpa (192.41.230.2)  0.289 ms  0.217 ms  0.209 ms
 2  r04chi-te-1-4-ptp-umich.ultralight.org (192.84.86.229)  5.734 ms  5.690 ms  5.691 ms
 3  192.84.86.10 (192.84.86.10)  5.683 ms !<10>  5.699 ms !<10>  5.644 ms !<10>

root@umfs11 ~# traceroute newmon.bnl.gov
traceroute to newmon.bnl.gov (130.199.3.7), 30 hops max, 46 byte packets
 1  vl4001-nile.aglt2.or.230.41.192.in-addr.arpa (192.41.230.2)  0.281 ms  0.215 ms  0.194 ms
 2  r04chi-te-1-4-ptp-umich.ultralight.org (192.84.86.229)  5.791 ms  5.738 ms  5.710 ms
 3  chi-ultralight.es.net (198.125.140.205)  6.021 ms  6.058 ms  6.056 ms
 4  chiccr1-starcr1.es.net (134.55.207.33)  6.215 ms  6.078 ms  6.041 ms
 5  clevcr1-ip-chiccr1.es.net (134.55.217.53)  15.141 ms  23.795 ms  15.043 ms
 6  washcr1-ip-clevcr1.es.net (134.55.222.58)  22.760 ms  22.723 ms  22.719 ms
 7  aofacr2-washcr1.es.net (134.55.218.78)  27.921 ms  27.919 ms  45.452 ms
 8  bnlmr1-aoacr1.es.net (134.55.217.57)  29.814 ms  29.872 ms  29.803 ms
 9  bnlsite-bnlmr1.es.net (198.124.216.178)  125.724 ms  31.603 ms  48.383 ms
10  newmon.bnl.gov (130.199.3.7)  29.663 ms  29.620 ms  29.615 ms

root@umfs11 ~# traceroute nettest10g.usatlas.bnl.gov
traceroute to nettest10g.usatlas.bnl.gov (192.12.15.25), 30 hops max, 46 byte packets
 1  vl4001-nile.aglt2.or.230.41.192.in-addr.arpa (192.41.230.2)  0.454 ms  0.225 ms  0.201 ms
 2  r04chi-te-1-4-ptp-umich.ultralight.org (192.84.86.229)  5.723 ms  5.707 ms  5.684 ms
 3  chi-ultralight.es.net (198.125.140.205)  5.997 ms  6.003 ms  5.952 ms
 4  chiccr1-starcr1.es.net (134.55.207.33)  6.073 ms  6.012 ms  6.039 ms
 5  clevcr1-ip-chiccr1.es.net (134.55.217.53)  15.084 ms  15.048 ms  15.051 ms
 6  washcr1-ip-clevcr1.es.net (134.55.222.58)  22.704 ms  22.704 ms  22.688 ms
 7  aofacr2-washcr1.es.net (134.55.218.78)  27.876 ms  27.907 ms  27.859 ms
 8  bnlmr1-aoacr1.es.net (134.55.217.57)  30.423 ms  30.169 ms  29.876 ms
 9  bnlsite-bnlmr1.es.net (198.124.216.178)  192.944 ms  204.639 ms  89.670 ms
10  nettest10g.usatlas.bnl.gov (192.12.15.25)  29.759 ms  29.733 ms  29.746 ms

[root@wood1-chi ~]# traceroute umfs11.aglt2.org
traceroute to umfs11.aglt2.org (192.41.230.31), 30 hops max, 40 byte packets
 1  umfs11.aglt2.org (192.41.230.31)  6.897 ms  6.961 ms  6.950 ms

[root@wood1-chi ~]# tracepath umfs11.aglt2.org
 1:  192.84.86.10 (192.84.86.10)                            0.055ms pmtu 9000
 1:  192.84.86.1 (192.84.86.1)                              3.318ms
 2:  192.84.86.230 (192.84.86.230)                          6.535ms
 3:  umfs11.aglt2.org (192.41.230.31)                       5.718ms reached
     Resume: pmtu 9000 hops 3 back 3

traceroute to newmon.bnl.gov (130.199.3.7), 30 hops max, 40 byte packets
 1  * * *
 2  * * *
 3  * * *
 4  * * *
 5  * washcr1-ip-clevcr1.es.net (134.55.222.58)  18.794 ms  18.859 ms
 6  aofacr2-washcr1.es.net (134.55.218.78)  23.969 ms  22.274 ms  22.281 ms
 7  bnlmr1-aoacr1.es.net (134.55.217.57)  24.317 ms  24.185 ms  24.261 ms
 8  bnlsite-bnlmr1.es.net (198.124.216.178)  24.235 ms  24.248 ms  24.249 ms
 9  newmon.bnl.gov (130.199.3.7)  24.038 ms  23.998 ms  24.034 ms

[root@wood1-chi ~]# tracepath newmon.bnl.gov
 1:  192.84.86.10 (192.84.86.10)                            0.061ms pmtu 9000
 1:  192.84.86.1 (192.84.86.1)                              3.172ms
 2:  chi-ultralight.es.net (198.125.140.205)              asymm  3   0.349ms
 3:  chiccr1-starcr1.es.net (134.55.207.33)               asymm  4   0.760ms
 4:  clevcr1-ip-chiccr1.es.net (134.55.217.53)            asymm  5   9.625ms
 5:  no reply
 6:  aofacr2-washcr1.es.net (134.55.218.78)               asymm  7  22.459ms
 7:  bnlmr1-aoacr1.es.net (134.55.217.57)                  25.035ms
 8:  bnlsite-bnlmr1.es.net (198.124.216.178)              102.907ms
 9:  bnlsite-bnlmr1.es.net (198.124.216.178)              asymm  8 103.916ms pmtu 1500
10:  newmon.bnl.gov (130.199.3.7)                         asymm  9  24.073ms reached
     Resume: pmtu 1500 hops 10 back 9

[root@wood1-chi ~]# traceroute nettest10g.usatlas.bnl.gov
traceroute to nettest10g.usatlas.bnl.gov (192.12.15.25), 30 hops max, 40 byte packets
 1  * * *
 2  * * *
 3  * * *
 4  * * *
 5  * washcr1-ip-clevcr1.es.net (134.55.222.58)  19.051 ms  19.115 ms
 6  aofacr2-washcr1.es.net (134.55.218.78)  24.143 ms  22.246 ms  22.250 ms
 7  bnlmr1-aoacr1.es.net (134.55.217.57)  24.642 ms  45.929 ms  46.031 ms
 8  bnlsite-bnlmr1.es.net (198.124.216.178)  24.518 ms  24.142 ms  24.155 ms
 9  nettest10g.usatlas.bnl.gov (192.12.15.25)  24.186 ms  24.144 ms  24.196 ms

[root@wood1-chi ~]# tracepath nettest10g.usatlas.bnl.gov
 1:  192.84.86.10 (192.84.86.10)                            0.054ms pmtu 9000
 1:  192.84.86.1 (192.84.86.1)                              2.888ms
 2:  chi-ultralight.es.net (198.125.140.205)              asymm  3   0.352ms
 3:  chiccr1-starcr1.es.net (134.55.207.33)               asymm  4   0.749ms
 4:  clevcr1-ip-chiccr1.es.net (134.55.217.53)            asymm  5   9.622ms
 5:  no reply
 6:  aofacr2-washcr1.es.net (134.55.218.78)               asymm  7  22.493ms
 7:  bnlmr1-aoacr1.es.net (134.55.217.57)                  25.018ms
 8:  bnlsite-bnlmr1.es.net (198.124.216.178)               40.501ms
 9:  bnlsite-bnlmr1.es.net (198.124.216.178)              asymm  8  73.872ms pmtu 1500
10:  nettest10g.usatlas.bnl.gov (192.12.15.25)             24.270ms reached
     Resume: pmtu 1500 hops 10 back 10

Information on NICs/stack settings:

  • wood1-chi.uslhcnet.org
[root@wood1-chi ~]# ethtool -i eth2
driver: myri10ge
version: 1.4.4
firmware-version: 1.4.38 -- 2008/12/18 23:27:11 m
bus-info: 0000:09:00.0

net.ipv4.udp_wmem_min = 4096
net.ipv4.udp_rmem_min = 4096
net.ipv4.udp_mem = 386208       514944  772416
net.ipv4.tcp_rmem = 4096        87380   16777216
net.ipv4.tcp_wmem = 4096        65536   16777216
net.ipv4.tcp_mem = 196608       262144  393216
net.ipv4.igmp_max_memberships = 20
net.core.optmem_max = 20480
net.core.rmem_default = 126976
net.core.wmem_default = 126976
net.core.rmem_max = 16777216
net.core.wmem_max = 16777216
  • umfs11.aglt2.org
root@umfs11 ~# ethtool -i eth2
driver: myri10ge
version: 1.4.0
firmware-version: 1.4.29 -- 2008/01/03 14:46:05 m
bus-info: 0000:0a:00.0

net.ipv4.tcp_rmem = 4096        87380   800000000
net.ipv4.tcp_wmem = 4096        65536   800000000
net.ipv4.tcp_mem = 786432       1048576 1572864
net.ipv4.igmp_max_memberships = 20
net.core.optmem_max = 20480
net.core.rmem_default = 135168
net.core.wmem_default = 135168
net.core.rmem_max = 800000000
net.core.wmem_max = 800000000

-- ShawnMcKee - 12 Feb 2009
Topic revision: r4 - 16 Oct 2009, TomRockwell
This site is powered by FoswikiCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding Foswiki? Send feedback