Anyone having issues with CloudFlare

Portal Home > Knowledgebase > Industry Announcements > Web Hosting Main Forums > Providers and Network Outages and Updates > Anyone having issues with CloudFlare

Posted by KMyers, 09-11-2011, 05:44 PM
Hello,
My blog has been up and down most of the day. I can access everything when I go to the server IP or direct.server.com, the problem is only when goinf through CloudFlare
Posted by panda4uhk, 09-11-2011, 08:19 PM
https://www.cloudflare.com/system-status.html

LAX had some upstream network issues this morning. Upstream is OK now.
Tweeted Sun Sep 11 19:30:46 +0000 2011
Posted by flam316, 09-11-2011, 10:46 PM
CloudFlare's Pingdom stats show no downtime, and I haven't experienced any downtime on any of my sites powered by CloudFlare. One of them also has Pingdom monitoring and I haven't gotten one downtime notification.
Posted by damoncloudflare, 09-12-2011, 05:09 PM
Hi,

While there were issues in LA briefly, I would recommend checking to make sure your don't have rules blocking our requests at the host or server level (behavior your described also leads to the situation you were having).

Our IPs:

• 204.93.240.0/24 (204.93.240.0 - 204.93.240.255)
• 204.93.177.0/24 (204.93.177.0 - 204.93.177.255)
• 199.27.128.0/21 (199.27.128.0 - 199.27.135.255)
• 173.245.48.0/20 (173.245.48.0 - 173.245.63.255)
• 103.22.200.0/22 (103.22.200.0 - 103.22.203.255)
• 141.101.64.0/18 (141.101.64.0 - 141.101.127.255)
Posted by pkhunter, 11-08-2011, 08:08 PM
When will you fix the CDN issues?

I am a "Pro" member. I am not availing of the free service.

Traceroute from Singapore, HK, Tokyo, Korea. All of them take a long way to the US servers (nlayer.net) and on average take 500ms.

This is UTTERLY USELESS.

Please fix your CDN. What's the point of having Asia data centers if no traffic is routed from them. My traceroutes have been using the key ISPs in these cities -- Starhub in Singapore, SoNet in Tokyo, SKT in Korea, and so on.
Posted by damoncloudflare, 11-08-2011, 08:16 PM
Hi Pkhunter,

You can contact us on the CloudFlare site with the IP addresses you did the traces (whaismyip.com) from and the traceroutes (we can have a network engineer take a look at it).

Asian routing is notoriously difficult. And while this is an overly simplistic explanation, the networks/service providers look for the most efficient path & may decide that routing elsewhere is better. This may not be an issue we can fix per se because it is out of our direct control...but having the IP address you did the trace from & the traceroutes does let us see if we can contact the provider for adjustments.
Posted by pkhunter, 11-08-2011, 08:24 PM
Yes, this has already been done several time since the last two months. The engineers have left it off as "we can only talk to the ISPs".

What does that mean? A proper CDN, if located in a proper local hub, does not need ISP conversations. Almost all other CDNs I have tried, including Softlayer's own (with INternap) work fantastically as they suggest.

This whole "most efficient path" nonsense does not hold much water if the same tripe is true from every single city, every single time. Don't tell me please that the most efficient path according to four leading ISPs in four leading Asian cities to cloudflare servers is via several hoops through US servers. The same ISPs think the most efficient path to Softlayer's CDN, for example, is within 6 hops with average of 35 ms. Compare that with 14 hops to Cloudflare with an average of 500 ms.

Recognize the problem instead of passing it off as some networking jargon. THat'll be a good start. Your network engineers haven't done squat. They have my threads and tickets with many traceroutes.
Posted by damoncloudflare, 11-08-2011, 09:13 PM
Hi,

I'm not blowing smoke when I tell you what they said is the issue. And we do pass along all issues relative to routing to our Network Engineers - so we're definitely not ignoring the issue at al. They are pretty familiar with what can/can't be done and/or what the limitations are in fixing some issues. I can assure you we will pass all the traceroutes along.
Posted by plumsauce, 11-08-2011, 11:47 PM

Quote:

Originally Posted by pkhunter

Traceroute from Singapore, HK, Tokyo, Korea. All of them take a long way to the US servers (nlayer.net) and on average take 500ms.

Straight ping from LAX to Singapore is usually right around 180ms on the network we use, and Nagano is 120ms. Even South Africa is only about 350ms.
Posted by cpoalmighty, 11-09-2011, 07:17 AM
I was thinking of putting one of my clients from the Asian part of the world on CloudFlare because our servers are located in the US and did not want to hamper their clients from reaching them, but I read some really negative things about cloudflare implementations and did not want to risk their site being "down" to visitors. Right now the speed is exceptional but nonetheless, I wanted a solution to save on their bandwidth usage.
Posted by damoncloudflare, 11-09-2011, 05:48 PM

Quote:

Originally Posted by cpoalmighty

I was thinking of putting one of my clients from the Asian part of the world on CloudFlare because our servers are located in the US and did not want to hamper their clients from reaching them, but I read some really negative things about cloudflare implementations and did not want to risk their site being "down" to visitors. Right now the speed is exceptional but nonetheless, I wanted a solution to save on their bandwidth usage.

Our uptime across our network is actually rock solid & easily exceeds 99.99+%. We can have brief service issues in a datacenter due to an attack, but this doesn't mean that all datacenters & visitors to your site would be having issues reaching the site.