CloudF[ai]l[u]are Lacks Failover Support

[drmike]

With all the recent site attacks elsewhere and continued finger pointing about CloudFlare, I went over there finally to check their stuff out.

Major problem, they have no support for failover.   Meaning no way to tell CF to point elsewhere when your backend server can't be reached.

Am I missing something about how to accomplish such an obvious thing with CloudFlare?  Absent such support, no wonder why things behind CF go down and stay down.

 

[XFS_Duke]

Yea, that is one of the things I've seen a lot. I'm a CloudFlare partner even though I'm not advertised on their site, not sure why but I'm asking soon, and I always thought that it should of been included in their plans. Possibly to send it to a StatusPage.IO page or something...

Which reminds me, sorry to thread jack, but any other services such as StatusPage.io available that is cheaper? lol

 

[drmike]

The failover piece is just entirely obvious.  It's one of the features I most use with reverse proxy logic in Nginx...  So conceptually the ability exists and isn't too hard to graft onto things.

One of their developers said about some DNS Zone hack ability, but that seems manual in nature.

API exists and is CURL friendly, so perhaps I'll monkey with that seeing if I can build something in testing to see such swap a downed backend with something else.

Guess I understand why sites go down even while CF is up front.... Not hard to punt a site offline when no failover logic there.

 

[Munzy]

You can specify multiple backend servers, and it will do very basic failover and round robin, but I mean "BASIC"

You can use there api and monitor your backend server and swap to failover, which is the best method.

Sorry about grammar. 

 

[AMDbuilder]

It would be nice if they supported failover, but until then solutions such as these get the job done.





@XFS_Duke I'm still looking for one, granted it might be easier to just build our own from the looks of things.

 

[drmike]

It just dawned on me why a certain site has a "cluster"... That is to say multiple A record frontends....  That's the hackaround to CF lacking failover...

 

[howardsl2]

You can achieve this using Monit. Works great for me. Please see:

DNS failover for cloudflare with monit
http://blog.booru.org/?p=12

Note that the DIUP API method for CloudFlare was deprecated a couple years ago. It is no longer in their documentation. Replace it with a currently supported method if you are concerned about this.

 

[XFS_Duke]

It would be nice if they supported failover, but until then solutions such as these get the job done.





@XFS_Duke I'm still looking for one, granted it might be easier to just build our own from the looks of things.

Yea, I was thinking about it...

 

[drmike]

You can achieve this using Monit. Works great for me. Please see:


DNS failover for cloudflare with monit

http://blog.booru.org/?p=12


Note that the DIUP API method for CloudFlare was deprecated a couple years ago. It is no longer in their documentation. Replace it with a currently supported method if you are concerned about this.

Is that your code?  Thank you if so... Good to see it and to inspire something if I don't just wholesale try to use it as-is.

That's entirely the direction I was headed!

 

[willie]

I don't use CF but I thought the basic idea with these services is your monitoring (run by you or them) tracks which of your servers is up, and pokes CF's API if some of your servers fail, to take them out of the load balancing mix.

 

[drmike]

I don't use CF but I thought the basic idea with these services is your monitoring (run by you or them) tracks which of your servers is up, and pokes CF's API if some of your servers fail, to take them out of the load balancing mix.

Indeed that's right.

Self rolling things is fine and all.  DIY route ads some complexity and costs.  Looking at 2 different severs minimum (3 really to be sane), then a fourth for the monitoring / API tie in.   Starts to add up.

I mean honestly, some monitoring and failover for $5-10 a month and I'd pay CF something maybe.

 

[mikho]

Is that your code?  Thank you if so... Good to see it and to inspire something if I don't just wholesale try to use it as-is.


That's entirely the direction I was headed!

A LES user wrote this http://www.lowendguide.com/3/monitoring/setup-a-ha-round-robin-blog-on-debian-7-with-cloudflare/


To add failover "security" for some cheap boxes.

 

[howardsl2]

Is that your code? Thank you if so... Good to see it and to inspire something if I don't just wholesale try to use it as-is.

That's entirely the direction I was headed!

Those are not my code, hehe. I learned this method when @INIZ mentioned it in an older thread on LET. The code (and blog) was written by someone else on the Internet.

 

[wlanboy]

Yup Cloudflare lacks some features.

They are pushing more into the DDoS protection business and lose sight of their core DNS business.

Just look at DNS Made Easy.

Add DNS Failover / System Monitoring Service $4.95 per A record per year.

 

[Steven F]

Yup Cloudflare lacks some features.

They are pushing more into the DDoS protection business and lose sight of their core DNS business.

Just look at DNS Made Easy.

Yeah, but your costs just went from $0.00 a year to $35 a year.

 

[drmike]

Yeah, but your costs just went from $0.00 a year to $35 a year.

I've always been a paid DNSMADEEASY ANNUAL customer.

The $4.95 per year per domain essentially for failover, it is well worth it and reasonably priced.

Fact is, I expect to pay something for a real service.  I am happy to actually --- within reason of course.