h.gtld-servers.net offline...
Unless I am mistaken, h.gtld-servers.net is offline and has been for an hour or two. I can't see the containing prefix, 192.54.112.0/24. http://www.ris.ripe.net/perl-risapp/prefixinuse.do?rrc_id=1000&Submit=Submit&.submit=type&sortby=time&outype=html&preftype=ematch&interval=1&prefix=192.54.112.0/24 Will
On 15-Jun-2006, at 09:41, Will Hargrave wrote:
Unless I am mistaken, h.gtld-servers.net is offline and has been for an hour or two. I can't see the containing prefix, 192.54.112.0/24.
I think you're mistaken about the server being off-line, since I can see it just fine from many places. The RIPE NCC dnsmon tool can also see it from its various probes: http://dnsmon.ripe.net/dns-servmon/server/?server=h.gtld- servers.net&show=SHOW I don't know any details about how that servers is deployed, however, so it's non-trivial to draw more conclusions about what problems you're having. Perhaps a single anycast node has some issues, or perhaps 192.54.112.0/24 has flapped a bit, and has been suppressed due to dampening in your neck of the woods. Joe
Joe Abley wrote:
I think you're mistaken about the server being off-line, since I can see it just fine from many places. The RIPE NCC dnsmon tool can also see it from its various probes: http://dnsmon.ripe.net/dns-servmon/server/?server=h.gtld-servers.net&show=SHOW
That's old data. This is a recent incident. (IIRC, the most recent data isn't publicly available). It's rolled over and you can see it going down in the live interface now: http://dnsmon.ripe.net/dns-servmon/server/plot?type=drops&server=h.gtld-servers.net&day=15&month=6&year=2006&hour=6&period=2h&shift=%2B6h
I don't know any details about how that servers is deployed, however, so it's non-trivial to draw more conclusions about what problems you're having. Perhaps a single anycast node has some issues, or perhaps 192.54.112.0/24 has flapped a bit, and has been suppressed due to dampening in your neck of the woods.
I did (and do) check on multiple ASs that I run and asked a few others to check, also checked looking glasses and so on. But anyway, it's back now, so nothing to see. Obviously a local problem of some sort. Will
On Thu, 15 Jun 2006, Will Hargrave wrote:
Joe Abley wrote:
I think you're mistaken about the server being off-line, since I can see it just fine from many places. The RIPE NCC dnsmon tool can also see it from its various probes: I did (and do) check on multiple ASs that I run and asked a few others to check, also checked looking glasses and so on.
So, out of curiousity, i loss of 1 of 13 gtld servers important? I believe (though I could be mistaken) that these are actually anycast as well. I think dropping even 2-3 of the servers probably wouldn't affect overall performance would it? won't bind pick the 'best' place to ask regardless and stick/prefer that over 'slow' servers? It's certainly an interesting data point, but how does it affect the network as a whole? (perhaps this was the 1 hour/year permitted for maint on the network/device in question?)
On Thu, 15 Jun 2006, Christopher L. Morrow wrote:
On Thu, 15 Jun 2006, Will Hargrave wrote:
Joe Abley wrote:
I think you're mistaken about the server being off-line, since I can see it just fine from many places. The RIPE NCC dnsmon tool can also see it from its various probes: I did (and do) check on multiple ASs that I run and asked a few others to check, also checked looking glasses and so on.
So, out of curiousity, i loss of 1 of 13 gtld servers important? I believe (though I could be mistaken) that these are actually anycast as well. I think dropping even 2-3 of the servers probably wouldn't affect overall performance would it? won't bind pick the 'best' place to ask regardless and stick/prefer that over 'slow' servers?
I think he was right to report it here - its operational issue with very large TLD. But it is certainly nothing to seriously worry about as dns compensates for such problems. -- William Leibzon Elan Networks william@elan.net
On Thu, 15 Jun 2006, william(at)elan.net wrote:
On Thu, 15 Jun 2006, Christopher L. Morrow wrote:
On Thu, 15 Jun 2006, Will Hargrave wrote:
Joe Abley wrote:
I think you're mistaken about the server being off-line, since I can see it just fine from many places. The RIPE NCC dnsmon tool can also see it from its various probes: I did (and do) check on multiple ASs that I run and asked a few others to check, also checked looking glasses and so on.
So, out of curiousity, i loss of 1 of 13 gtld servers important? I believe (though I could be mistaken) that these are actually anycast as well. I think dropping even 2-3 of the servers probably wouldn't affect overall performance would it? won't bind pick the 'best' place to ask regardless and stick/prefer that over 'slow' servers?
I think he was right to report it here - its operational issue with very large TLD. But it is certainly nothing to seriously worry about as dns compensates for such problems.
Ooops, sorry, I didn't mean to harangue Will for reporting it, I was asking if it was in fact not a big deal because the system has compensation methods in place to deal with even 2-4 device outages. sorry for the confusion.
participants (4)
-
Christopher L. Morrow
-
Joe Abley
-
Will Hargrave
-
william(at)elan.net