Return Of The DNS Bugs

July 17, 2012 12:11 PM

I've tried deleting cookies, but I'm still getting DNS issues connecting to the blue today; it's been buggy all day. (Kind of like with this MeTa a year ago.) Am I the only one? (Hope not, because I'm at work....)
posted by EmpressCallipygos to Bugs at 12:11 PM (14 comments total)

When you say "DNS issues" can you be more specific and let us know what else you've tried...

- another DNS configuration?
- another computer?
- another ISP?

Is it only happening on metafilter.com or on the subsites as well? Can you generally be a lot more specific about what is happening?
posted by jessamyn (staff) at 12:15 PM on July 17, 2012

I don't know how to try other DNS configurations, and I'm not able to do much as I'm at work and can't move to another computer/try another ISP.

What I mean is: I am reading a thread on the blue, and I will attempt to go to the main page by clicking on the MetaFilter link and I"ll get a dns_unresolved_hostname error. I also get similar dns errors trying to click on my MeMail box. Projects comes up fine, as does AskMe. But I was also having this problem earlier today and fixed it by shutting down and rebooting my computer; everything worked fine for a few hours and then it started again.

The earlier thread I link to is actually a very good description of what I've been experiencing in terms of problem; I admit that I can't do much in terms of trying different solutions, which I'll admit is not helpful. (If I'm the only one facing this today I'll suck it up.)
posted by EmpressCallipygos at 12:24 PM on July 17, 2012

Oh - my own "recent activity" link also gives me a DNS error. AskMe doesn't.
posted by EmpressCallipygos at 12:26 PM on July 17, 2012

Been behaving for me. Just another data point.
posted by cjorgensen at 12:29 PM on July 17, 2012

It has been working fine for me all day as well, and I've been a bit meta ocd today because it's too damn hot to go outside.

I suspect it's your local isp, your local network, or whatever dns server your work network is using.
posted by HuronBob at 2:40 PM on July 17, 2012

Yeah, for the record we haven't heard from anyone else about problems via the contact form either.
posted by cortex (staff) at 3:07 PM on July 17, 2012

I re-started my computer earlier and it fixed it again, and now I'm at home and having no trouble.

I would have no compunctions whatsoever about this thread being closed and left to stand as a sort of Internet Stupid Tax or something.
posted by EmpressCallipygos at 3:08 PM on July 17, 2012

I think Matt paid the "Internet Stupid Tax" with that cat-scan post.
posted by HuronBob at 3:13 PM on July 17, 2012

My picture is in the dictionary next to "Internet Stupid Tax Reasons"
posted by AugustWest at 4:07 PM on July 17, 2012

I was getting a DNS error briefly this morning. It lasted like five minutes and things have been normal since.
posted by hoyland at 4:45 PM on July 17, 2012

Sounds like you were probably having some problem specific to the domain name server you are using — either problems it was having itself, or problems with your connection to it.

What's interesting to me, though, is why your computer isn't caching those lookup at all — I've been out of the loop for a long, long time so maybe I've forgotten what I once knew. But I really thought that DNS lookups were normally cached by the client.
posted by Ivan Fyodorovich at 5:18 PM on July 17, 2012

Well, it looks like Metafilter's DNS records have a really short TTL, only about 15 minutes. So if you're having some intermittent problem resolving mefi's address, then every 15 minutes you have another chance for the problem to occur. Multiply that by the number of subsites and add gremlins?
posted by hattifattener at 11:49 PM on July 17, 2012

This test page points out the SOA RETRY is too high (three hours). If easydns is using a standard zone transfer to keep secondary nameservers in sync, a name server that fails to update would stay out of sync for that long. That said, I don't see how stale data would be a problem right now, and I suppose the problem is closer to your ISP.
posted by Tobu at 1:01 AM on July 18, 2012

This test page points out the SOA RETRY is too high (three hours).

Sigh.

Back in they day, you wanted short SOA RETRY and REFRESH values to keep your name servers in sync. Modern implementations have DNS notify (RFC 1996) and don't want secondary name servers hammering on them every 10 to 30 minutes like they did in the days where retry was 300 and refresh was 1800 seconds.

The big advantage of notify is you can have very quick DNS update propagation without having to have your secondaries beat up your primary by updating every five minutes. With DNS Notify in play, you want retry on the order of 1-4 hours, and refresh on the order of days. Notify allows your master NS to tell the slaves to refresh now, which means your changes go out much quicker, and the SOA Refresh/Retry are just a backup to the notify scheme.

However, DNS analyzers can't tell if DNS Notify is in place. So, they assume not and report that you're out of compliance with the RFC 1912 suggestions.

The value for SOA RETRY is fine -- though the parent server should have a NS record for dns0.easydns.com, since the SOA record has that server as the master name server for the domain.
posted by eriko at 12:04 PM on July 18, 2012