This was the final summary post for the NANOG ISP Y2K mailing list. Special thanks to Merit for allowing the ad-hoc group of Internet Service Providers to share the name for the NANOG ISP Y2K event. This post is a couple of hours early, because I want to go to sleep. NANOG ISP Y2K Summary Status 2-JAN-2000 12:30 UTC Y2K Summary Although the "hotlines" got a lot of press, much of the information came via e-mail. It was difficult to dial into the international conference bridges near midnight in various countries due to congestion on the international voice circuits. The Internet is a dynamic place. All numbers may vary depending on where you are on the Internet. But the trends should be the same. The Internet global routing table showed a slow decline in networks from 71,200 on December 25, 1999 to 70,911 on December 30, 1999. On December 31, 1999 the networks steadily declined to 69,812 by 1 hour before Midnight EST. During this hour the routing table quickly dropped to 67,795 routes, and held steady from Midnight until 5am EST. After 5am EST January 1, 2000 the number of networks started increasing steadily to 70,455 on January 2, 2000. The number of unique ASNs in the global route table declined from 6,380 on December 30, 1999 to 6,330 at January 1, 2000 Midnight EST. The number of terminating ASNs showed a similar decline from 5,180 to 5,130. Which leads to the conclusion transit networks did not disconnect during the rollover. Peak traffic levels across MAE-East were lower on Friday December 31, about 1.6Gb versus 1.9Gb for a normal workday. There was a slight but noticable 0.2Gb dip in traffic at Midnight EST. Followed by a slight increase after Midnight EST. All well below peak traffic capacity. Traffic flow followed the normal sine wave pattern, except for the dip at Midnight EST. Summary of ISPs 2 ISPs reported a routing problem, isolated to normal circuit problems. 1 ISP reported voice-call (POTS) congestion in New Zealand shortly after midnight through several voice carriers. Internet backbone connectivity was not affected. This was the pattern through most time zones. Several ISPs reported trouble synchronizing NTP servers. Congestion on the NIST network is believed the cause. NTP servers elsewhere did not show any problems. 1 ISP reported messages sent in elm showed year as 100 instead of 2000. Patch was available prior to Y2K. Several reports of user-written scripts containing poor date handling practices. Report of one web site defacement after close of business. Report of some unauthorized domain name transfers at the beginning of the holiday weekend. Not unusual, seems to happen every holiday weekend. Two ccTLDs were down for part of the rollover. 13 ccTLDs had partial problems. No Y2K faults were found. Circuit problems, administrators turning off servers, and ordinary DNS problems. Exchange points worldwide reported a decrease in traffic immediately before their midnight local time. A sharp increase immediately after midnight local time, but less than peak capacity. Certain "event" web sites saw traffic increase 20 times over normal levels. Some event sites report being overwhelmed. But overall, traffic appears to have been lower than peak business day capacity. The use of many country-specific information sites, and the rolling nature of the Y2K event helped distribute the traffic worldwide. In general public interest waned about 30 minutes after local Midnight in each timezone, leaving more bandwidth for each successive timezone. Midnight US Eastern time did dominate changes in traffic and routing. A few rumors sprung up throughout the night. Most were quickly determined to be unfounded. But overall, not as many as I thought would spread. Issues known/predicted prior to the Y2K rollover Some sites plan to shutdown or disconnect from the Internet over the New Year's weekend Some certificate authority certificates expire on December 31, 1999 Increase in voice and cellular calls immediately around local midnight may cause some congestion on circuits Information from the media (Note: ISP names are from publicly announced information. Names will not be included from non-public sources.) C I Host (www.cihost.com) issued a press release about a name server problem. No client data files were damaged. The data corruption that occurred Dec. 29 is isolated to the [company local] nameservers only and the restoration remains unrelated to any Y2K issues. Keynote Internet Performance Update #2 reports slower access to New Zealand after rollover, 6.2 seconds. No information what the measurement was before the rollover. Appears to be localized to specific web sites (i.e. congestion) France's National Weather Service Internet site had a display problem with a date. The page shows 01/01/19100 instead of 01/01/2000. United States Naval Observatory web time site had a display problem with a date. The page shows 19100 instead of 2000. Hacker target Japan Y2K Center. No intrusion or damage reported. Several stories based on Keynote press releases report the Internet passed through the Y2K rollover unaffected, other than a few spots of congestion around particular web sites. The www.y2k.gov web servers were predicted able to handle 40 million hits. Only 3 million showed on January 1. Many, MANY, ISPs, web hosting, and access providers issued press releases throughout January 1, 2000 announcing they had no Y2K problems (and I think they all sent me a copy). Summary from the Y2KCC/JP The number of reported troubles 25 Troubles with Stratum1 NTP server 4 Trouble with ICQ 1 Troubles with ccTLD 15 small partial troubles 13 others 2 Trouble with router 1 Trouble with Y2KCC/JP system 1 Latency of a particular site 1 Suspect of cracking 1 Trouble with NNTP server 1