Hello everyone,
The College of DuPage has a NOAAport receiver with a Novra S300N, and two
Ubuntu boxes as ingestors, so that if one fails, the other will still provide
data. To accomplish this, they use an enterprise level network switch, a
Nortel/Avaya 5520, that costs a lot of money (and not the $50 gigabit switches
you can get from Newegg, or wherever), to split the data feed so that both
servers can get and ultimately relay the data.
When trying to receive the test GOES-16 imagery months ago, it wasn't coming
across their NOAAport feed. I thought it was a glitch with their Novra box, or
the network switch. But now that GOES-16 data is being sent, neither NOAAPort
ingest server were getting the GOES-16 data on the NOTHER feed. What could be
wrong?
We restarted the LDM on both servers, deleting and remaking the queues. That
didn't do anything.
We then double checked the ldmd.conf files, which had a few minor issues, and
also checked the logs,
of course. Still, no GOES-16 data was showing up.
We then did a packet sniff from one of the servers to the switch. No GOES-16
test data was being sent to the boxes from the switch!
We then made sure the Ubuntu kernel parameters were set correctly, per the LDM
instructions. One box didn't have it correct; the other was correct. We
rebooted both boxes after doing checks. That didn't get us GOES-16 data, either.
So then, today, I convinced a College of DuPage meteorology program employee to
go out to the bitter cold dish farm and climate-controlled "dog house" where
the Novra and the servers are. As he power-cycled the Novra, I was doing an
ldmadmin watch on the NOTHER feed on the ingest servers. When the Novra came
back up after the power cycke...no GOES-16 data was coming across to the
servers!
So then we looked to the network switch. The employee grabbed a jumper cable,
and ran an Ethernet cable directly from the Novra to one of the two ingest
servers. Then, finally, GOES-16 data started pouring in on that server!
But why? How can a $1,000+ enterprise level switch have a problem like this?
Well, as it turns out, it wasn't the fault of the switch at all. See:
224.0.1.[1-10] have reserved internet protocols on them:
http://www.networksorcery.com/enp/protocol/ip/multicast.htm
For those of you who don't like to understandably click on links from strangers
(and geeky meteorologists):
INTERNETWORK
CONTROL BLOCK
224.0.1.0
VMTP<http://www.networksorcery.com/enp/protocol/vmtp.htm> Managers.
RFC 1045
224.0.1.1
NTP<http://www.networksorcery.com/enp/protocol/ntp.htm>, Network Time Protocol.
RFC 1119<http://www.networksorcery.com/enp/rfc/rfc1119.pdf>
224.0.1.2
SGI-Dogfight.
224.0.1.3
Rwhod.
224.0.1.4
VNP.
224.0.1.5
Artificial Horizons - Aviator.
224.0.1.6
NSS, Name Service Server.
224.0.1.7
AUDIONEWS - Audio News Multicast.
224.0.1.8
SUN NIS+ Information Service.
224.0.1.9
MTP<http://www.networksorcery.com/enp/protocol/mtp.htm>, Multicast Transport
Protocol.
RFC 1301<http://www.networksorcery.com/enp/rfc/rfc1301.txt>
224.0.1.10
IETF-1-LOW-AUDIO.
Uh huh. When the NWS picked those IP addresses to encapsulate the data, they
picked IP numbers that are used and reserved for enterprise-level switches.
As for the College of DuPage? I just got this email from head network engineer,
Dave Bukowski, who came up with this solution: you need to enable unknown
multicast flooding on those IP's. Here's what Dave did, and his email with the
solution:
-----------
FOUND IT!!!!
This was in the config
show running-config
.....SNIP.....
!
! *** IGMP ***
!
vlan igmp unknown-mcast-allow-flood 224.0.1.1
vlan igmp unknown-mcast-allow-flood 224.0.1.2
vlan igmp unknown-mcast-allow-flood 224.0.1.3
vlan igmp unknown-mcast-allow-flood 224.0.1.4
vlan igmp unknown-mcast-allow-flood 224.0.1.5
vlan igmp unknown-mcast-no-flood enable
.....ENDSNIP.....
Added
vlan igmp unknown-mcast-allow-flood 224.0.1.6
vlan igmp unknown-mcast-allow-flood 224.0.1.7
vlan igmp unknown-mcast-allow-flood 224.0.1.8
vlan igmp unknown-mcast-allow-flood 224.0.1.9
vlan igmp unknown-mcast-allow-flood 224.0.1.10
vlan igmp unknown-mcast-allow-flood 224.0.1.201
So now:
SDBDF01#show vlan igmp unknown-mcast-allow-flood
Allowed Multicast Addresses
-----------------------------------------------------------------------
224.0.1.1 224.0.1.2 224.0.1.3 224.0.1.4
224.0.1.5 224.0.1.6 224.0.1.7 224.0.1.8
224.0.1.9 224.0.1.10 224.0.1.201
Total Multicast MAC Addresses: 0
Total Multicast IP Addresses: 11
Now we have a flood of data coming in on both noaaport1 and noaaport2. It was
multicast and we were blocking the flood, but had to specifically allow the
flooding by multicast address.
-Dave
Thanks, Dave. We spent at least 10 hours trying to figure this out. On a $50
switch, you might not have a problem, but if you do decide to have multiple
ingesters by going through a network switch...if you don't get all of the data,
your reception may be fine...it's the switch that's been programmed to stop the
flow of data.
Thanks to Tom Yoksas as he helped us try to figure this mess out, to Mike
Zuranski for going out in the bitter cold and spending hours on the phone as we
tried to figure this out, and for Dave Bukowski who ran the ball for the
touchdown after we discovered it was the switch blocking the data.
Gilbert
Gilbert Sebenste
Staff Meteorologist
Environmental Health and Safety
Labs for Wellness 154 | DeKalb, Illinois 60115
815-753-5492
gilbert@xxxxxxx<mailto:gilbert@xxxxxxx>
http://weather.admin.niu.edu<http://weather.admin.niu.edu/>
Everyone. Home. Safely.
[NIU]