I’m working on lemmy-meter which is a simple observability solution for Lemmy end-users like me, to be able to check the health of a few endpoints of their favourite instance in a visually pleasing way.
👉 You can check out a screenshot of the pre-release landing page.
💡 Currently, lemmy-meter sends 33 HTTP GET requests per minute to a given instance.
For a few reasons, I don’t wish lemmy-meter to cause any unwanted extra load on Lemmy instances.
As such I’d like it be an opt-in solution, ie a given instance’s admin(s) should decide whether they want their instance to be included in lemmy-meter’s reports.
❓ Now, assuming I’ve got a list of instances to begin w/, what’s the best way to reach out to the admins wrt lemmy-meter?
PS: The idea occurred to me after a discussion RE momentary outages.
Interesting, FYI there is a similar project: https://lestat.org/
Thanks for the link. Had no idea about that.
33 HTTP GET requests per minute to a given instance.
That is way beyond acceptable use, and would likely have your service blocked. There exists these services too :
https://lemmy.fediverse.observer/stats
Maybe those do what you’re trying to do?
There is not an “admin inbox” for lemmy instances. You can hit the endpoint
/api/v3/sitefor information about an instance including the admins list.beyond acceptable use
Since literally every aspect of lemmy-meter is configurable per instance, I’m not worried about that 😎 The admins can tell me what’s the frequency/number they’re comfortable w/ and I can reconfigure the solution.
You can hit the endpoint /api/v3/site for information about an instance including the admins list.
Exactly what I was looking for. Thanks very much 🙏
I’m not worried about that 😎
You should be. Your name will be associated with abuse forevermore.
The admins can tell me what’s the frequency/number they’re comfortable w/ and I can reconfigure the solution.
Or you can set some sane defaults and a timeout period. 1 request / 5 mins is fine to check if something is online and responding.
sane defaults and a timeout period
I agree. This makes more sense.
Your name will be associated with abuse forevermore.
I was going to ignore your reply as a 🧌 given it’s an opt-in service for HTTP monitoring. But then you had a good point on the next line!
Let’s use such important labels where they actually make sense 🙂
Are these 33 different requests or do you hit the same endpoint multiple times?
I’d probably default to every 5 minutes at most, but I guess if it’s up to the admin then it’s all good. 33 requests per minute shouldn’t be a ton of load if it’s all read requests.
Speaking as an admin of an instance here, 33 requests a minute is not “all good”.
Not without asking, but if the admin is okay with it then sure. I don’t see the point of any sort of monitoring making that many requests per minute though.
Indeed. IIUC, OP said 33 reqs/min is a ceiling and tunable on a per-target basis.
If the target is a Cloudflare instance, you could perhaps even do 300 reqs/min without even being noticed.
That is way beyond acceptable use
Each server would have its own acceptable use policy. Also consider the social detriment of Cloudflare nodes. We could even say @bahmanm@lemmy.ml has a moral /duty/ to overwork the Cloudflare nodes :)
@Blaze@discuss.tchncs.de: thanks for pointing out lestat. That’s one of the very few services of this kind to be responsible enough to red-flag the Cloudflare nodes. I hope @bahmanm@lemmy.ml follows that example; though it could still be improved on.
It’s misleading for any tor-blocking Cloudflare node to have a 100% availability stat just because by design it deliberately breaks availability to a number of users in an arbitrarily discriminatory fashion. https://lemmy-status.org/ and https://lemmy.fediverse.observer/stats do a bit of a disservice by not omitting or flagging #Cloudflare nodes.
Why does it need to make 33 requests per minute? Surely the data doesn’t have to be that up to date?
Agreed. It was a mix of too ambitious standards for up-to-date data and poor configuration on my side.
Update 1
Thanks all for your feedback 🙏 I think everybody made a valid point that the OOTB configuration of 33 requests/min was quite useless and we can do better than that.
I reconfigured timeouts and probes and tuned it down to 4 HTTP GET requests/minute out of the box - see the configuration for details.
🌐 A pre-release version is available at lemmy-meter.info.
For the moment, it only probes the test instances
I’d very much appreciate your further thoughts and feedback.