tcs@BRL.MIL (Terry Slattery, SECAD) (01/12/89)
Hi,
We've just seen some very strange behavior from some PSNs on the milnet and
would like some suggestions on how to proceed from here.
Synopsis:
BRL internal hosts have been able to to talk to ARDEC.ARMY.MIL
(ARDEC.ARPA, addr 26.1.0.45) in the past. However, for the last week or so,
the connections to them have been horrible to this site. We can ping
26.0.0.45 (a TAC) with good turn-around times (< 500ms), but cannot ping
ardec at all. The only other host on that PSN that we can contact is
TCACCIS-BAY.ARPA (26.2.0.45, also < 500ms). Testing from other milnet hosts
(usna.mil) shows that they can reach ardec, but that the round trip times
are vary quite widely (500ms - 5 sec). Pings to the two working node 45
hosts also exhibit a long round trip time for the first few packets and then
stabilize at < 500ms. Pings to other hosts around the milnet show no such
behavior. Routes from both BRL-GATEWAY and USNA are identical. Ardec can
occasionally reach BRL hosts, but we can not reach them. In the recent
past, we have had problems maintaining connections to them. Host
connections via other PSNs are fine.
Could there be a PSN or line problem which would cause this type of
behavior? Talking to the Milnet CONUS NOC was fruitless ("Maybe your
terminals are incompatible; your machine is running UTX and they are
running BSD.") Milnet NOC was able to telnet to both hosts so they claim
the problem is with one of us (also suggesting that we checking host
tables, etc.)
Anyone have any suggestions?
-tcsstjohns@BEAST.DDN.MIL (Mike St. Johns) (01/13/89)
Hmm... Hey Dave (Mills)! You got your satellite hunter tool handy? Seriously, I'll ask the PSN gurus to take a look at recent performance stats. Do you have any data on direct pings from your gateway to these hosts? (As opposed to sliding through the gateway from one of your internal hosts?) After a quick look at the NIC records, my guess is that a substantial number of those hosts are X.25 basic service customers. But, you should be able to ping at least the first 4 ports (all 1822). Mike
malis@BBN.COM (Andy Malis) (01/24/89)
Terry, I haven't seen a follow-up on the TCP-IP list since your original message (I just got back from vacation) and I'm curious if you are still experiencing the problem with ARDEC.ARPA. I have a ping running right now and the average time is running around 380 ms. I'm getting about the same average for 26.0.0.45 as well (this is through two gateways between the BBN internal net and the MILNET). Experiencing a longer round trip time for the first ping or two is to be expected; the PSNs are in the process of setting up a subnet end-to-end virtual circuit. Once the connection has been established, data flows. The behavior you described (two hosts at the same PSN having widely differing operating characteristics) usually points to problems with an individual host: either a problem in the host itself, or a hardware problem with its access line or port (on either end of the line). The folks at the CONUS NOC are usually very good at solving this sort of problem, and are assisted by development folks up here in Cambridge when necessary. I suggest giving them another try if this is still an ongoing problem. If that fails, mail to tcp-ip certainly gets attention, since it is read by both folks at the DDN PMO and in BBNCC. Regards, Andy Malis Manager, BBNCC PSN Development Support
tcs@BRL.MIL (Terry Slattery, SECAD) (01/24/89)
Thanks for your inquiry. Yes, things have gotten much better in the last week. There have been problems with our PSN which may have been part of the problem. The symptoms of the PSN problem is messages from the PSN to our gateway which indicate insufficient resources in our PSN for connections to some sites. A trouble ticket has been issued in Mike Muuss' name for this problem. It is still open at this time and we are still getting these advisory packets from the PSN. The behavior I noted in my message to the list was not due to problems with a specific host (or so it seemed). Both had proper routes to the other and could be pinged from a third host (usna.mil). These pings had reasonable round trip times. The aberrant behavior was only between ardec.arpa and hosts at BRL. We could occasionally get packets through by starting a pinging from ardec.arpa. Anyway, the problem I reported seems to have been resolved since ping round trips are now reasonable (< 500ms). Our PSN is also being reloaded tonight with new software, presumably to fix the problem with insufficient resources. -tcs