Load balancer single point of failure?

[smozely]

Hi,

Been looking at cloudfoundry and it looks great. One question though, it seems like the Apache frontend server would be a single point of failure. Is there anything or anything in the works that could remove this as a single failure point? or am I wrong to think its a single point

I do understand why it would be less risk than an app server or db server if it was a single POF, but I have to answer so RFP documentation where a system is asked for that doesn't have a single point of failure.

Any feedback appreciated.
Steve

[ceracm]

Steve,

Thanks for your question.

You are right in thinking that the Apache server is a SPOF.
But since it is a cloud, if the instance fails CF can relaunch a replacement quite quickly - you wouldn't have to wait, for example, for the hardware vendor to ship a new motherboard. And, so the downtime would be minimal.

However, a better solution would be to run multiple Apache servers behind, for example, an elastic load balancer.We thought about providing that as a deployment option but we can't commit to a date.

I hope this helps.

Chris


Chris

[smozely]

Thanks Chris,

Yes that does help,
I didn't think of advertising how quickly we can start another one when an instance fails.

Hopefully I can convince my client that this isn't an issue either.

Just for my info not using the ELB a side effect of using sticky sessions in the Apache config?

Steve

[ceracm]

CloudTools.org/CloudFoundry.com predate ELB.

Many Java applications need sticky session routing.
ELB very recently introduced support for routing HTTP requests so for some Java applications you could dispense with Apache (for routing)
But ELB still can't route HTTPS requests so you would need Apache or the equivalent for that.

[smozely]

Cool, thanks for the info Chris

[coldblooded1]

Yes! I finally found the answer to my question. Chris, you're my god Thanks again for the info.

[dylanbarrell]

Would love a solution to this as it would give me more security in the case of failures as well as help to scale beyond what a single Apache machine can support.