DigitalOcean Unveil Floating IPs

[perennate]

@HBAndreiWe are discussing technology no emotions here - I guess my point was clear however I do respect your opinion but overall its a lame solution - I saw an article also sharing the same opnion with some details on whyhttps://z0z0.me/2015/10/21/why-digitalocean-floating-is-fake/

As a current customer of DO - and many others my point is : They could have done better.

Thanks,
VPSFan

That blog post is laughably ignorant. Floating IP is a commonly used term, and Digital Ocean's implementation of it is virtually identical to that used by Amazon (elastic IP), OpenStack, and others. Just because they haven't launched load balancer as a service system doesn't make the floating IP system any less useful.

It's like when a company says they're going to start selling boxes, people start complaining that they're not selling shoes. wtf?

Edit: but the funniest part is, they provide three guides on how to use the system to achieve automatic failover, and people still complain:

* https://www.digitalocean.com/community/tutorials/how-to-create-a-high-availability-setup-with-corosync-pacemaker-and-floating-ips-on-ubuntu-14-04

* https://www.digitalocean.com/community/tutorials/how-to-set-up-highly-available-web-servers-with-keepalived-and-floating-ips-on-ubuntu-14-04

* https://www.digitalocean.com/community/tutorials/how-to-create-a-high-availability-setup-with-heartbeat-and-floating-ips-on-ubuntu-14-04

 

[Petertk]

When someone wants to use floating IP the main idea is when a node goes down for various reasons the cluster to detect the unavalability of the node and to move the IP address to the other nodes with the related services with less downtime as possible. With DigitalOcean's Floating IP the node cannot need to do constant probes which has to be scripted by the user itself then if the scripts would detect a node down it would need to send a request to the DigitalOcean API which would change it's NAT1:1 policy from the down node to the active node. Therefore all these steps would generate a downtime in services, much higher then with simple clustering solution. Not talking about the fact that with scripting you cannot do a probe so often as the cluster software it does it itself. 

Regards,

 

[HN-Matt]

Floating IP is a commonly used term, and Digital Ocean's implementation of it is virtually identical to that used by Amazon (elastic IP), OpenStack, and others. Just because they haven't launched load balancer as a service system doesn't make the floating IP system any less useful.

It's like when a company says they're going to start selling boxes, people start complaining that they're not selling shoes. wtf?

Edit: but the funniest part is, they provide three guides on how to use the system to achieve automatic failover, and people still complain:

* https://www.digitalocean.com/community/tutorials/how-to-create-a-high-availability-setup-with-corosync-pacemaker-and-floating-ips-on-ubuntu-14-04

* https://www.digitalocean.com/community/tutorials/how-to-set-up-highly-available-web-servers-with-keepalived-and-floating-ips-on-ubuntu-14-04

* https://www.digitalocean.com/community/tutorials/how-to-create-a-high-availability-setup-with-heartbeat-and-floating-ips-on-ubuntu-14-04

I've never had a VPS with DO and have no dog in this fight, but yeah, that's sort of what I was getting at. Seems to be a learning curve thing, or something that caters to different levels of skill and experience, maybe? I can see how some will want to cut to the chase from square one, but others who are new to, or not as familiar with, concepts like 'floating IPs' or 'automatic failover' may want to learn more about it at a slower pace before diving headlong into the ready-made automation cluster.

tl;dr DO, you have failed to impress the floating IP connoisseurs, guess it's time to throw the towel in.

 

[perennate]

When someone wants to use floating IP the main idea is when a node goes down for various reasons the cluster to detect the unavalability of the node and to move the IP address to the other nodes with the related services with less downtime as possible.

I see floating IP addresses and load balancing as a service systems as mostly orthogonal features.

• Floating IP address: allows you to manually move IP between instances, or automatically failover with your own control software. For example, if you are performing upgrades to your web application, you may want to test the upgrade on another server before exposing the changes publicly. With floating IP, you can commit the change just by moving the IP address to the new server.
• Load balancing as a service (LBaaS): when you want to balance requests across multiple machines and have the system automatically mark nodes as failed (and don't care to have any control over this process, beyond a reasonable set of configurable parameters). In this case, the external IP isn't associated with any particular instance, but rather with a pool of instances that requests may be forwarded to.

I concede that it would be cool if Digital Ocean supported LBaaS, but I don't think it makes sense to complain that they don't support it in the context of their release of floating IPs. A floating IP is by nature associated with an instance; if you want automatic failover, then this association shouldn't exist, and a separate system should deliver that functionality. I also concede that Digital Ocean shouldn't be marketing floating IP's primary purpose as automatic failover, since there are so many other (more mundane) areas where it can be applied.

Regardless, the complexity of following DO's tutorial on automatic failover is negligible compared to the complexity of setting up primary-backup replication that would be needed behind the scenes to support this failover.

Also, I'm especially confused by the blog post mentioned earlier because of how it talks about failover between datacenters. It seems that they want some kind of anycast system for virtual machines, which again is separate from both LBaaS and floating IP addresses. Anycast would be nice (although there's a huge amount of complexity around coping with flapping routes in the middle of a TCP session), but that doesn't make the deployment of floating IP functionality any less useful.