The results are in, after seeing about 800 boosts on the post to see how the Mastodon embed actually affects our server load. Here's what happened....basically nothing for us.
I think itsfoss doesn't have proper caching set up for their CDN and just gets bombarded with every request for the site, possibly every subsequent request by every user as well.
@chris as a dynamic site too, we also don't exactly use a plain html cache either, but our website is also around 1mb without browser cache on an article, there's is like 11mb!
it's kinda funny that instead of hiring someone to submit a patch for their issue, #itsfoss decided to publish the article about their technical problems
The article reads superficial, like vague request for commercial proposal but for free lol
> #Mastodon instance admins, please fix our computer, or else we'll stop reposting to your social network
@graphito "just hire someone" is not really a valid thing here though for smaller sites, the article framing was poor given how heavy their site is but the issue behind it does exist
@graphito It is an issue, especially for self-hosters. I enjoy watching all the victim blaming and denigration that goes along with it, because people pointing out issues with an open source protocol _must_ mean they want to wall garden the entire internet.
There’s a lot missing here. How many CPU cores? Is this normalized load? What was the latency? Were there timeouts? Why are you looking at ~10 minute intervals when the reported problem is with spikes of ~1 minute? Very “works on my machine” type post
Sorry, didn’t mean to be mean. I was concerned that you were using your platform to show that this isn’t a problem without understanding what the problem is.
When a link is posted on an account that’s federated to a lot of instances, it gets accessed many times all at once to generate link previews. That means you should see very short bursts of traffic, and that can be enough to cause requests to queue up and possibly time out or get rejected. Hope that helps
I guess I’m trying to say it might not have been fine, because it might not show up as a high load average. If you have server logs (especially proxy logs e.g. nginx) from the time you posted the link, you might see what I’m talking about. The concern I have is that people will see this and think it’s evidence there isn’t a problem given the reach your posts are getting.
Liam @ GamingOnLinux 🐧🎮
in reply to Liam @ GamingOnLinux 🐧🎮 • • •Christopher Snowhill
in reply to Liam @ GamingOnLinux 🐧🎮 • • •Liam @ GamingOnLinux 🐧🎮
in reply to Christopher Snowhill • • •Christopher Snowhill
in reply to Liam @ GamingOnLinux 🐧🎮 • • •Christopher Snowhill
in reply to Christopher Snowhill • • •Liam @ GamingOnLinux 🐧🎮
in reply to Christopher Snowhill • • •Daniele Pantaleo 🦥:verified:
in reply to Liam @ GamingOnLinux 🐧🎮 • • •Wheatly
in reply to Liam @ GamingOnLinux 🐧🎮 • • •it's kinda funny that instead of hiring someone to submit a patch for their issue, #itsfoss decided to publish the article about their technical problems
The article reads superficial, like vague request for commercial proposal but for free lol
> #Mastodon instance admins, please fix our computer, or else we'll stop reposting to your social network
Folks, this is not journalism...
Liam @ GamingOnLinux 🐧🎮
in reply to Wheatly • • •Baggypants
in reply to Liam @ GamingOnLinux 🐧🎮 • • •Wheatly
in reply to Liam @ GamingOnLinux 🐧🎮 • • •I admit, similar approach might be justified for small websites (1-2 people) who cannot have budget to deal with technical problems
But surely #itsfoss have the budget to deal with tech. Instead they turned to public activism to save a bit of it.
Why not publish the tech description on freelancer sites and have the patch submitted by monday for $300?
Turns out, to rile up the audience and instance admins is simply cheaper 😳
cc @Baggypants
Grampa
in reply to Liam @ GamingOnLinux 🐧🎮 • • •Liam @ GamingOnLinux 🐧🎮
in reply to Grampa • • •Grampa
in reply to Liam @ GamingOnLinux 🐧🎮 • • •Sorry, didn’t mean to be mean. I was concerned that you were using your platform to show that this isn’t a problem without understanding what the problem is.
When a link is posted on an account that’s federated to a lot of instances, it gets accessed many times all at once to generate link previews. That means you should see very short bursts of traffic, and that can be enough to cause requests to queue up and possibly time out or get rejected. Hope that helps
Liam @ GamingOnLinux 🐧🎮
in reply to Grampa • • •Grampa
in reply to Liam @ GamingOnLinux 🐧🎮 • • •