Technical (internet) help PLEASE...

[markem]

Originally Posted by Dood

OK, this is wierd.

72.249.0.66 is the last router before the website.
Don't know why it isn't passing some requests.

Here is the owner information for that router. The contact info is
Phone: +1-214-630-3100 27 (Office)
Email: support@colo4dallas.com

NetRange: 72.249.0.0 - 72.249.191.255
CIDR: 72.249.0.0/17, 72.249.128.0/18
NetName: COLO4-BLK2
NetHandle: NET-72-249-0-0-1
Parent: NET-72-0-0-0-0
NetType: Direct Allocation
NameServer: NS1.COLO4DALLAS.NET
NameServer: NS2.COLO4DALLAS.NET
Comment:
RegDate: 2006-08-25
Updated: 2007-10-17

OrgAbuseHandle: CAM9-ARIN
OrgAbuseName: Colo4Dallas Abuse Manager
OrgAbusePhone: +1-214-630-3100
OrgAbuseEmail: abuse@colo4dallas.com

OrgNOCHandle: NOC1718-ARIN
OrgNOCName: NOC
OrgNOCPhone: +1-214-630-3100
OrgNOCEmail: support@colo4dallas.com

OrgTechHandle: CAR47-ARIN
OrgTechName: Colo4Dallas ARIN Requests
OrgTechPhone: +1-214-630-3100

OrgTechEmail: arin@colo4dallas.com

The network operations center may be able to help
Name: NOC
Handle: NOC1718-ARIN
Company: Colo4Dallas
Address: 3000 Irving Blvd
City: Dallas
StateProv: TX
PostalCode: 75247
Country: US
Comment:
RegDate: 2005-01-25
Updated: 2007-02-12
Phone: +1-214-630-3100 27 (Office)
Email: support@colo4dallas.com

[smitdavi]

Thanks Mark...I'd owe ya big if we can figure this out.

[markem]

Do any of the admins know if the server has more than on network interface? They would all have the same IP address, since only only one is assigned to the domain name.

Also, if you call the NOC (network operations center), have them check their DNS to see if there is more than one IP address associated with the name internal to them. Also have them check the ARP cache to make sure that an invalid entry isn't stuck.

[smitdavi]

Originally Posted by vstrommark

Do any of the admins know if the server has more than on network interface? They would all have the same IP address, since only only one is assigned to the domain name.

Also, if you call the NOC (network operations center), have them check their DNS to see if there is more than one IP address associated with the name internal to them. Also have them check the ARP cache to make sure that an invalid entry isn't stuck.

Mark do you need anything from me?

[markem]

Originally Posted by smitdavi

Mark do you need anything from me?

Not at this time, David. The problem looks to be very local. It is probably inside the co-lo facility where the server is housed.

[smitdavi]

Originally Posted by vstrommark

Not at this time, David. The problem looks to be very local. It is probably inside the co-lo facility where the server is housed.

Alright...let me know if you guys need any info or need for me to do anything on my end!

[mmblz]

Originally Posted by smitdavi

Alright...let me know if you guys need any info or need for me to do anything on my end!

see if it works next time you have a chance...

Originally Posted by vstrommark

Not at this time, David. The problem looks to be very local. It is probably inside the co-lo facility where the server is housed.

Most Data centers will have two pipes or more, for redundancy, say if they're in Texas, having one go to Chicago and the other to LA or NY.... So as possible as it is, because a large amount of people still can get to it, I'd still think to look at the ISP end first. Users should contact their ISP only if having problems because if not, they'll show it as being fine. Now the admins should alert the web host as well to a problem, so they can investigate, but something like this unless you're on the inside isn't as easy to track down.

Basing this off the experience I have with working at an ISP previously and with out the networks my company and it's customers use, which are scattered across the US, and several other countries...having some locations store different servers for Apps and such.

[markem]

Originally Posted by Hardcz

Most Data centers will have two pipes or more, for redundancy, say if they're in Texas, having one go to Chicago and the other to LA or NY.... So as possible as it is, because a large amount of people still can get to it, I'd still think to look at the ISP end first. Users should contact their ISP only if having problems because if not, they'll show it as being fine. Now the admins should alert the web host as well to a problem, so they can investigate, but something like this unless you're on the inside isn't as easy to track down.

Basing this off the experience I have with working at an ISP previously and with out the networks my company and it's customers use, which are scattered across the US, and several other countries...having some locations store different servers for Apps and such.

Agreed, more or less. Since the problem is manifesting itself at the 1st upstream router, the problem, statistically, is local to the facility. That is, unless they do physical layer shunting across the US, which is very costly and make little or no sense.

The problem could be a myriad of things from a bad network interface on the server to a HW or SW issue on the upstream device (router or switch) or internal DNS configuration issue or similar.

There is always the possibility that the server is running IP chains or something similar and that the config is screwed up in some way, but that's way far fetched base on what I know.

Like you, I have load of network engineering experience plus telecommunications OS design, work on networking standards committees, and a prior life as a network programmer and college professor. That doesn't mean that either of us knows squat about this problem, however.

Colo4dallas is a good company. (have used them myself)
Try emailing them and ask them why the routing errors, maybe on their end (since the hops do get to dallas)

Best thing you can do is contact the internet provider, if you're having problems so they can investigate. The way things should work is your connection goes to say comcast, and they connect to ATT/Verizon via a frame relay cloud, which picks the route to the CA server, wherever it is, and if it's not routing correctly, they can modify the route or find out if there's a problem. Most likely there's an outage somewhere that's causing the page to be unavailable. Something you could try is using an anonymous proxy server in your internet browser, which basically routes your traffic from their location to here, which assuming it's not on the same nodes, will find a route that's not bad.

[smitdavi]

Originally Posted by Hardcz

Best thing you can do is contact the internet provider, if you're having problems so they can investigate. The way things should work is your connection goes to say comcast, and they connect to ATT/Verizon via a frame relay cloud, which picks the route to the CA server, wherever it is, and if it's not routing correctly, they can modify the route or find out if there's a problem. Most likely there's an outage somewhere that's causing the page to be unavailable. Something you could try is using an anonymous proxy server in your internet browser, which basically routes your traffic from their location to here, which assuming it's not on the same nodes, will find a route that's not bad.

So is that something that I need to do. Call AT&T?

[markem]

Originally Posted by smitdavi

So is that something that I need to do. Call AT&T?

No, because it isn't your problem, it is CAs problem. Someone from CA would need to contact them or their account rep.

[mmblz]

Originally Posted by vstrommark

Do any of the admins know if the server has more than on network interface? They would all have the same IP address, since only only one is assigned to the domain name.

yes, it has a number of network interfaces, with different ip addresses (afaik)

i've told them about the problem and they couldn't find any problem with configuration.

[markem]

Originally Posted by mmblz

yes, it has a number of network interfaces, with different ip addresses (afaik)

i've told them about the problem and they couldn't find any problem with configuration.

You really want to talk to someone who can log on to the router right before the server and do some testing. That seems to be where the problem is at.

My BGP table is bigger than your BGP table