Alt servers

[nathanchu]

The main github readme stats server went down with 402 Payment Required today (as you probably know). This has happened multiple times and I have a proposed solution. I'd be happy to host an alternate server for this purpose, assuming it doesn't get too much traffic. I'd like to hear your thoughts on this @anuraghazra, and whether anyone else would consider hosting a community alt server.

[anuraghazra]

@nathanchu hmm thats a really nice suggestion.

So what do you propose? we will gather few people who are willing to deploy github-readme-stats to their vercel instance and we will use those URLs to use as alternatives?

[anuraghazra]

no yeah i got that we will have something like that.

but i did not get your point actually.
What would this github action will do? thats my question.

[nathanchu]

It would prevent it going offline but also it would be easier to deploy because you wouldn't need to generate a token, you would only need that.

[nathanchu]

It would generate the svg image and save to a specified path

[anuraghazra]

Ahhh okay okay got it. so instead of being served dynamically from the serverless function you are saying we can just convert this to a github action and people can use that too.

[nathanchu]

I've set up a simple implementation: https://github.com/nathanchu/github-readme-stats-action. Here's the demo: https://github.com/nathanchu/github-readme-stats-action-demo

[kingthorin]

So a week or two ago I setup my own Vercel instance because my card wasn't updating properly (likely because of the traffic that the author's instance was getting). It's cool, it works. But every few days when I visit my page the image doesn't load, if I reload the page it pops. Do Vercel instances go to sleep or something when they have super low traffic? I'm starting to wonder if @nathanchu was on to something with the idea of a GitHub workflow. I'm thinking to simply update a static image once or twice a day or something. (My stats don't need to be accurate up to the minute.)

cache_seconds - set the cache header manually (min: 1800, max: 86400)

I'm also wondering if I should bump this way up. Looking in dev tools the image seems to be coming through cached for 2h (7200sec).

[kingthorin]

I think the issue I'm seeing is that the CDN (camo) isn't keeping it available because I have near zero profile traffic. I'm going to look into either just updating the image once or twice a day and storing/accessing it statically (via raw.github) or an hourly job that link checks or crawls just my profile to keep images/links fresh in the CDN (camo).

[kingthorin]

So in the end here's what I did to keep a fresh image available. (I only tackled it a few hours ago so it's still PoC).

1. Add a README.md to my "special" profile repo's .github/workflows directory that has the same image tag as my profile README.md (so that visiting it will refresh the image/cache but not increment my hit counter).
2. Setup a workflow that visits/takes a screenshot of the page every 90mins.
3. Attach the screenshot to the workflow run, so that I can go back and make sure it did actually load 😉 for any given run.

[kingthorin]

I think I got all the quirks of the workflow sorted out now.

[kingthorin]

I decided to go a different way with this. Instead of trying to do a workflow to maintain activity/cache I'm just going to statically get and push the svg. kingthorin/kingthorin@17c0f61

[rickstaa]

Closing this as the GitHub action is tracked in #2179