Thorny DNS problems

yahoo_moe

Mods, hard to find the exact right place for this so hopefully this is it. Apologies if not.

Problem: The website I manage is periodically down - but not for everyone at the same time (i.e. it might be accessible in the US or via a UPC connection here, while it won't be accessible via our eircom connection, say).

Details

* We switched our hosting from one company to another about 8 weeks ago, but the registration is still with the old host.
* We launched a new site about the same time - in fact, as soon as the new host's nameservers were in control of the domain.
* After the switch, the old host had retained an old C-NAME record. They deleted this over 4 weeks ago though.
* When it's down, ping/nslookup fail to find the host or claim a non-existant domain. Generally the non-www version of the URL could be pinged (see below) but essentially there seems to be some error resolving the DNS.
* Looking at our new host's error logs, I can see that the requests aren't getting as far as their servers. They've also tested this for me.
* Last week, I copied the site to a new webspace on the current host's servers (shared hosting btw) with a new IP. The problems reappeared this morning.
* The old site had only been accessible through the www.domain.com URL but the new one was set up to be accessible through both. When the site was down, it used to still be partly accessible through the non-www domain (never associated with the old IP address). i.e. it found the new nav and displayed it almost in text form but couldn't pull the CSS, images, etc. from the www domain. This wasn't the case this morning though.
* I mentioned ISPs above but only as an example - the problem isn't limited to eircom. It's not limited to our office network either.
* Our office LAN (or at least some machines) were connecting to the web using eircom's old DNS servers. This has now been updated and obviously the problem reappeared this morning.
* Our email still goes through the old host.

Any ideas?

E39MSport

Not a domain expert (no pun intended) but I experienced something similar before where one proxy server had cached old ip addresses resulting in my browser periodically failing to see a url for a migrated service.


edit - just saw that its not limited to your office network. Soz.

Dr.Silly

tried flushing DNS ?

yahoo_moe

Thanks for the replies.

Dr. Silly, yup - started off thinking it was a DNS caching issue on our local network so flushed DNS locally on the machine and at server level, as well as restarting router, etc. Still sporadic problems (maybe just as well given it's obviously not a solution that scales up well).

E39MSport - yes, not just an internal problem unfortunately.

Any other thoughts?

Dr.Silly

have you tried a tracert to see where the problem may be ?

yahoo_moe

Yeah was running tracert early on but it wasn't really telling me a lot.

It basically just mirrored the ping results / page loading: it traced the route to the non-www version of the URL fine when that was loading and pingable, but wouldn't run on the www version when it's down (or non-www if that's down too).

I've also tried various dig tests and generally it's fine from the server of whatever site I'm using (robtex, kloth.net) but times out if I dig via the eircom nameservers.

Dr.Silly

yahoo_moe wrote: »

Yeah was running tracert early on but it wasn't really telling me a lot.

It basically just mirrored the ping results / page loading: it traced the route to the non-www version of the URL fine when that was loading and pingable, but wouldn't run on the www version when it's down (or non-www if that's down too).

I've also tried various dig tests and generally it's fine from the server of whatever site I'm using (robtex, kloth.net) but times out if I dig via the eircom nameservers.

what about hard coding the hosts file ?

yahoo_moe

I'm 90% sure that didn't make a difference when I tried it.

Maybe our IT company has set the hosts file so that it doesn't override the server-level entries?

qbasic

Its hard to make out your setup, this is what I understood from your OP.

You have a domain say, web1.com which was previously hosted by another provider they also maintained the DNS domain.

You moved web1.com site to a new provider leaving the DNS domain hosting with the old provider. At the same time that you moved web1.com to the new provider you set up a new site web2.com and the new provider hosts the DNS domain.

You periodically have issues getting to WWW.web1.com however no issues getting to web1.com. You also have mail on web1.com, do you have problems with your mail for this domain?

Dr.Silly

yahoo_moe wrote: »

I'm 90% sure that didn't make a difference when I tried it.

Maybe our IT company has set the hosts file so that it doesn't override the server-level entries?

possibly. worth a shot.

so you MX records are obviously still pointing to the old hosting company.
Are you 100% sure that that's the only communcation with them ?

yahoo_moe

qbasic wrote: »

You have a domain say, www.web1.com which was previously hosted by another provider they also maintained the DNS domain.

You moved web1.com site to a new provider leaving the DNS domain hosting with the old provider, leaving the domain registration with the old provider but moving DNS control to the new provider. At the same time that you moved web1.com to the new provider you set up a new site web2.com and the new provider hosts the DNS domain updated web1.com with completely new content.

You periodically have issues getting to WWW.web1.com however mostly no issues getting to web1.com. As of yesterday (the first time the problems had appeared since switching web1.com to a new webspace on the new provider's servers, this may not be the case anymore - neither domain was accessible. You also have mail on web1.com, routed through the old provider's servers with no problems

Sorry, not quite - edited in the quote above.

yahoo_moe

Dr.Silly wrote: »

so you MX records are obviously still pointing to the old hosting company.
Are you 100% sure that that's the only communcation with them ?

Not 100% unfortunately as I can't see their systems - it's all phone-based with no web CP.

Initially this was one of the problems - old C-NAME records were showing up pointing to the old host's nameservers, and an SOA record named them at one point too. In several phone calls though, they assured me that they no longer hold any of those old records for the domain, and they're no longer showing up in dig tests, etc. Also, these old records were showing up when digging through the old, no-longer-maintained eircom DNS servers that our LAN was using at that point.

Dr.Silly

yahoo_moe wrote: »

Sorry, not quite - edited in the quote above.

Are you allowed to give us the URL's, if not, if you want to PM ?

yahoo_moe

I'll send a PM now.

Sposs

Make sure your hosting company have deleted all records of the site from their servers.

Sounds like eircom with the fact they don't give you a cp, so might take a couple of phone calls before it actually gets done.

yahoo_moe

It's not actually eircom but yes, the lack of a CP is a pain.

It took several calls but then I was assured that they no longer held anything other than the reg. The next couple of times it was acting up after that I got onto them again and they said the same.

One possible solution is to move DNS control back to them - the servers that picked up the original change to the new provider should pick that up too and the ones that occasionally don't know where to look might stop being confused?

It'd be a bit of a pain losing control of things a bit but it'd probably be worth it for a stable site...

mneylon

When it's not visible who is impacted? ie. is it just you / your office?

If it's just your office I'd suspect an issue with your resolver

mneylon

Sposs wrote: »

Make sure your hosting company have deleted all records of the site from their servers.

That won't make any difference as he's talking about DNS resolution not email

It's the important difference between authorative and recursive resolvers etc.,

AndrewMc

Try entering your domain name into http://www.zonecut.net/dns/ and check how your DNS setup looks to the outside world.

yahoo_moe

Blacknight wrote: »

When it's not visible who is impacted? ie. is it just you / your office?

If it's just your office I'd suspect an issue with your resolver

No it's more widespread than that. There have been issues in Canada, though those seem to have disappeared now, but also for a number of other people in Ireland, across several ISPs.

They're all on office networks to the best of my knowledge but it's a pretty small sample size so I'm not reading much into that.

yahoo_moe

AndrewMc wrote: »

Try entering your domain name into http://www.zonecut.net/dns/ and check how your DNS setup looks to the outside world.

It's showing the right nameservers for the non-www domain, and not finding nameservers for the www version of the URL. However, that just seems to be how that engine works, as it did the same when I checked for rte.ie and www.rte.ie.

I've checked the site's setup using several different services like that and it's always normal, even when the site's 'down' for me.

yahoo_moe

Thanks to everyone for their help with this, should be sorted now - AndrewMc seems to have nailed down the problem (1 of the nameservers for .ie TLD was retaining old DNS records).

Thanks to Dr. Silly as well for his help.

Trojan

yahoo_moe wrote: »

Thanks to everyone for their help with this, should be sorted now - AndrewMc seems to have nailed down the problem (1 of the nameservers for .ie TLD was retaining old DNS records).

Glad to hear you got it resolved.