> Here, “a Node.js server” refers to one Node.js server. A single, unique instance of a Node.js server (with or without Docker) with no horizontal scaling and no zero-downtime deploys is not a viable deployment strategy for serious projects.
Ah, the penny drops. The idea that you can’t run a traditional server and must rely on serverless vendor if you’re “serious”, and that NextJs is not convenient to Netlify, is at the core of this blog, camouflaged by concerns of openness. I’m most certainly agree with openness, and quite honestly I dislike NextJs and Vercel, but suggesting that side stepping lock in altogether by simplifying down to traditional techniques is not “serious” makes me bristle a little, especially when you say we can’t do something that we have been doing for many, many years without your platform.
> Ah, the penny drops. The idea that you can’t run a traditional server and must rely on serverless vendor if you’re “serious”
That's not at all how you should read this. They later on give an example of exactly what kinds of problems you'll run into once you start needing to horizontally scale you Next.js servers (e.g. as pods in k8s, which is not serverless):
> The issue of stale data is trickier than it seems. For example, as each node has its own cache, if you use revalidatePath in your server action or route handler code, that code would run on just one of your nodes that happens to process that action/route, and only purge the cache for that node.
Seeing as a Node.js server running Next.js serving SSR or ISR (otherwise you'd just serve static files, which I personally prefer) is not known to have the greatest performance, you will quickly run into the need of needing to scale up your application once you hit any meaningful amount of traffic.
You can then try to keep scaling vertically to avoid the horizontal pains, but even that has limits seeing as Node.js is single-threaded, and will run into issues with the templating part of stringing together HTML simply taking too long (that is, compute will always block, only I/O can be yielded).
The common solution for this in Python, Ruby, and JS/Node.js is to run more instances of your program. Could be on the same machine still, but voila! you are now in horizontal scaling land, and will run into the cache issues mentioned above.
There was not really anything in the article that should have lead you to believe that this was a "serverless only" issue, so I think the bashing against Netlify here is quite unwarranted.
> There was not really anything in the article that should have lead you to believe that this was a "serverless only" issue, so I think the bashing against Netlify here is quite unwarranted.
It's not because you can use an external cache like Redis[1]. You can scale to hundreds of instances with an external redis cache and you'll be fine. The problem is that you can't operate on Netlify scale with a simple implementation like that. Netlify can't afford running a redis instance for every NextJS application without significantly cutting into their margins (not just from compute cost, but running and managing millions of redis instances at scale won't work).
Clearly Vercel has their own in-house cache service that they have priced in their model. Netlify could run a redis instance per application, though more realistically it needs its own implementation of a multi-tenant caching service that is secure, can scale, cost effective, and fits their operational model. They are not willing to invest in that.
Interesting and definitely something platforms must take into consideration.
Now back to the post, implementing custom cache is not something Netlify is strongly complaining about. They are mostly asking for some documentation with rather stable APIs. Other Frameworks seems to provide that.
> Netlify could run a redis instance per application, though more realistically it needs its own implementation of a multi-tenant caching service that is secure, can scale, cost effective, and fits their operational model. They are not willing to invest in that.
But they have done that, as they say in the post.
Disclosure: used to work at Netlify, now work at Astro
The ability to self-host on serverless compute is just a subset of the challenges. Even with your own VPS or ECS setup, I think it's legitimate to say a serious project requires server redundancy, and as soon as you have more than one running server it opens up requirements to synchronize cache state, which is a significant challenge. This has nothing to do with serverless architecture.
It's their whole business model, to convince developers that their walled garden is the only viable option. I've met a lot of newish developers that believe it too.
> suggesting that side stepping lock in altogether by simplifying down to traditional techniques is not “serious” makes me bristle a little
This is a strawman. You're misinterpreting the word "serious". They are using it to mean scalable, not about unimportance/ability. At some point in the scaling process, it will be more effective to scale to another machine than stay on a single one at which point you need a lot of other primitives like the article mentions. E.g. a shared cache with proper invalidation mechanisms. If you dont need scale, then you're right, you dont have to worry about this. I will also note that it is slightly odd to use a framework like next.js if you arent (or planning on) running at scale because most of its features (e.g. SSR) are entirely performance oriented. Essentially, the whole point of the article is that despite being "open source" you cannot run next.js at scale yourself without a massive investment of your own.
> Essentially, the whole point of the article is that despite being "open source" you cannot run next.js at scale yourself without a massive investment of your own.
I don't know about that. Asking a service provider to provide an implementation for a cache interface isn't a "massive investment". It's an investment sure, but it's the type of investment that seems should be customizable per provider depending on their needs, technologies they want to bet on, etc. It seems to me the problem with Netlify isn't comfortable putting in the investment to have a NextJS specific cache service. It's understandable considering they don't control the framework and to them it's just another option, so they don't want to invest in it too much.
(disclaimer: Netlify employee)
The big challenge with the cache interface atm is not using Redis (personally, I love Redis). It's that this interface is far from being a straightforward GET/SET/DELETE. Rather, you need to learn the default implementation and all its nuances (for different payload types), and duplicate/transform all the logic.
The division of labor between what the framework does and what the platform developers (or any other developer working on a high-scale/high-availability deployment) need to do, has to be fixed. If this happens - plus better docs - you should be able to "just use" Redis.
"Fortunately, thanks to Vercel’s outreach, we’ve established lines of communication with the Next.js team. With this in place we hope to gain better insights into upcoming changes to the framework."
So instead of having it in open, Netlify now have their own channel of communication with Vercel. This still continue to feel like a closed garden and makes it difficult for small vendors to have proper NextJS hosting.
(Disclosure: I am running DollarDeploy which allows hosting NextJS (amongst others) on your own server. And we don't use docker!)
(disclaimer: I work at Netlify)
It's not a 1:1 channel, we're pushing for as much work in the open as possible, and extending the OpenNext group to more members.
IMHO one major challenge is the amount of work one has to do to build an adapter, if we can make the spec leaner and move logic out of e.g. the CacheHandler, that's gonna make it easier for you to create your own (or you could state that some features/modes are not supported - but for that, you'd need to have a very good idea of what these even are)
I changed the example to point to latest NextJs 14 and fixed some cache-handler bugs it works fine, and I deploy it successfully with DollarDeploy to DO VPS https://github.com/huksley/next-cache-handler
This is a great point and definitely something we're mindful of. That sentence actually ended with "and share these with the community" in the draft, but we can't make promises on behalf of the Next.js team.
Our hope is that the OpenNext initiative's growth and connections with the Next.js team is just a first step toward shifting some Next.js governance out to the open. The article calls out that the OpenNext members are hoping to contribute to public RFCs in the future, with Next.js core team collaboration.
Unfortunately commercial companies tend to work behind closed doors/gardens. We are working on Low-Ops.com and would love to fully support NextJS. Self-hosting is the future :)
Worked at Netlify briefly, and the amount of plates they have to spin to keep all these integrations so smooth was super impressive. Some super impressive stuff under the hood. Also a lot of coordination between different projects. Shame to see Vercel not assisting with these efforts as other frameworks do.
What's shocking to me is that they deploy "hundreds of thousands" of self-hosted apps? I wonder what percentage of overall NextJS apps are on Vercel, and what percentage aren't.
I think NextJS is still a decent framework to use, so long as you're not going into any of their RSC or other "full stack" features and use it with the "export" flag (which just generates HTML/CSS/JS that you can upload anywhere).
One thing that is very annoying is that their next/image component still does not support "export" and only works if you deploy to Vercel. Considering that Gatsby figured this out so early on is just shameful. Granted there are a couple of 3rd party packages that add proper image support like Gatsby had, it's still very annoying and quite frankly sad because we all know this is a business decision to keep folks locked into Vercel.
Ah, the penny drops. The idea that you can’t run a traditional server and must rely on serverless vendor if you’re “serious”, and that NextJs is not convenient to Netlify, is at the core of this blog, camouflaged by concerns of openness. I’m most certainly agree with openness, and quite honestly I dislike NextJs and Vercel, but suggesting that side stepping lock in altogether by simplifying down to traditional techniques is not “serious” makes me bristle a little, especially when you say we can’t do something that we have been doing for many, many years without your platform.