Thoughts on hosting scalable Jitsi on DigitalOcean

Is anyone here hosting a large scale Jitsi deployment on DigitalOcean? If so, any chance you could share your thoughts/gotchas/advice on whether it is a suitable platform?

Also curious about whether meet.jit.si/jaas considered DigitalOcean when planning move out of AWS, and why Oracle Cloud was chosen. Was it down to region availability?

As far as I can tell, beyond 10TB egress, prices are cheaper on DO compared to OCI. We’ve done a trial migration to OCI and frankly the experience was not great compared to what we’re used to in AWS. Considering a DO trial, but hoping to hear from others with more experience with DO before committing to this effort.

Thanks.

I have no experince with OCI but I have been using DO for a long time (AWS too). DO is my first choice if I don’t need an AWS specific service or if the customers don’t force me to use AWS

More of a company initiative and contracts negotiation than anything else.

Now I know who to bother with n00b questions about DO :stuck_out_tongue:

Quick one, I don’t see any facility for autoscaling in DO unless you go down the k8s cluster route. How do you autoscale your JVB instances in DO? Do you use custom scripts/services to handle autoscaling using DO APIs?

I have no active autoscaling Jitsi on DO now but yes, you need to build your own autoscaling solution using API. JMS monitors active usage and can create new JVB using API.

Depending on the size/load of the cluster, using permanent droplets may cost lower than an autoscaling system on AWS.

This might be helpful:

That’s a good point.

I can probably get away with fixed scaling in the short term, but what I’d miss from ASGs is also the High Availability aspect of it – set the number of instance you want running, then have it auto-recover if instances die or fail healthchecks. Easy enough to automate that with custom code, but it all adds to the effort estimates :slight_smile:

Cheers Freddie.

Aaron did point that out when I was scratching my head on how to handle graceful scale-in on OCI. And there’s a PR out to make it work for DO too. It’s good stuff, but not quite something we can plug into our current infra without further work.

Not discounting it, for sure. I’m at the point where I’m scouting out all possible migration paths out of AWS and weighting out operational costs vs build cost.

Too many options is always better than no options :slight_smile: