PostHog
Docs: explain how to scale PostHog for self hosted users
Context
We have several large enterprises with high volumes approaching us.
We do not think we should focus primarily on these customers (ie provide tons of bespoke support), as that'll come at the cost of our core product quality for thousands of other companies. They bring a ton of revenue, but we have built a remarkable machine for mid market growth that needs optimization first.
We've seen some extraordinarily high open source instances in the wild (we had one with 2billion events in the last 5 days), so users can be successful if they have an SRE working on it.
We should do two things:
- In sales, tell customers that must self host "hey get the OS version up and running, that's your first step". If they can't do that, then they're not worth pursuing (unless they can then use cloud instead), if they can, they'll probably be 10 times easier to deal with
- In website and docs, we explain how to scale things better than we do today, so that it's easier for SREs at other companies to succeed without us
Proposal
I'd suggest @pjhul drives (and gets some help from @guidoiaquinti and/or @fuziontech), to cover:
- Which services we have
- How to get them approximately right in the first place
- How to monitor them
- What to do when ie Zookeeper goes down, or Clickhouse goes down, or (insert other service here)
- (This is my layperson's guess at what we need, feel free to change this list)
Feels like this is something we could support with some comms and marketing content as well.
- Once the content is up, create a cohort of users to whom we think this information would be useful in the future and send an email comm to let them know about this asset so they never need to ask for support.
- Feels like there could be a HN-y piece in this: 'What we learned about scaling on-prem deployments to X billion event scale' @andyvan-ph
- A case study could support this message for future customers and help them understand what's involved with self-hosting so they make the right choice. I was looking at Netdata, but if there's another candidate that better fits this topic then let me know.
What we learned about scaling on-prem deployments to X billion event scale
Or 'Why we decided to make Enterprise customers self-serve (and how it went)'
Feels like this is something we could support with some comms and marketing content as well.
(it does create some risk of people moving to OS from cloud, thus hurting revenue, so we probably ought to be careful with the above until we've also got a page explain that self hosting is never cheaper than cloud)
Suggest we move all of the above into a separate issue btw - totally valid ideas
Definitely agree with this and happy to drive! I'll start doing some planning on what information I need to put together. Next week I can try and set up some meetings with some of the engineers to start collating some of the fixes they've put together for large accounts in the past and also what we do on our end to address these challenges.
