📣 NEW: I realize we’re still in a “code freeze” for Yarn.social’s backend yarnd
; however I wanted to share a new feature that just got built this weekend.
This means the following two experimental features are not complementary and can hopefully both be promoted soon™ 🤞
- moving_average_feed_refresh
- websub
I would encourage anyone that normally runs yarnd
from the main
branch or the prologic/yarnd:latest Docker Image to give this a try and report any issues you come across!
You can enable this feature websub
or both websub
and moving_average_feed_refresh
in your Manage Pod UI and entering the features there one per line and
hitting “Save”.
Thank you! 🙇♂️
@david@netbros.com My apologies @david@netbros.com 🙇♂️ I should have done a better job here in documenting the new feature (maybe that’s why it’s still feature gated? 🤣) – Nevertheless…
what are the visible subtleties for users?
Hopefully none. The _only noticeable difference for cross-pod users in the Yarn.social network is real-time(ish) feed/cache refreshes as User A on Pod X makes a Twt and User B on Pod Y sees it within ~5s or so.
To see timeline, discovery, and mention changes now one must manually reload pages. Are these features changing this?
No, this will likely never change with the SSR (Server-Side Rendered) yarnd
Web Interface as-is. We may (I’ve been talking about it for a while now) introduce a “You have X new items, Click here to refresh” banner at the top at some point when @ullarah@txt.quisquiliae.com and I figure out how to build that.
How does this affects
--fetch-interval
Basically, fetching “subscribed” feeds will be done via WebSub. During the Cache fetch cycle, which is controlled by --fetch-interval
, subscribed feeds are skipped.
This has the effect of reducing inbound and outbound load on pods via “polling” and instead pods are notified in near real-time of a feed on a Pod being updated (note this is a yarnd
only feature).
@david@netbros.com Wow! 😳 High praise 🥳 Thanks @david@netbros.com 🙇♀️
One thing I’d like to show and point out is the effect this feature has had on my Pod alone:
The dip back up was me turning it off to fix some bugs 😅
@mckinley@twtxt.net Thanks bud 🤗
@screem@yarn.yarnpods.com How did you “update”? There is no new binary release yet… You either have to “build from source” or run the latest Docker image prologic/yarnd:latest
– If you do the later, you might have to docker pull prologic/yarnd
first 😅
@screem@yarn.yarnpods.com Well done 🥳 I assume you’ve now enabled websub
too 😂
@news@twtxt.net I will try later today maybe tomorrow and provide some feedback if I find out, what these tings do.
If you guys don’t know, there is a handy docker-image-extractor script which I am using https://github.com/jjlin/docker-image-extract/blob/main/docker-image-extract because on uberspace I cannot run docker so I have to take what’s inside the image and run that. Example can bee seen here https://lab.uberspace.de/guide_vaultwarden.html
It’s really nice to see the CPU usage reduced by quite a lot (especially on my pod, which is quite a bit busier)
Has anyone else noticed drops in CPU usage on their Pods since enabling websub
? 🤔