I thinking of uninstalling the excellent #Wordpress #ActivityPub Plugin and defederating my sites from the #Fediverse.
-
I thinking of uninstalling the excellent #Wordpress #ActivityPub Plugin and defederating my sites from the #Fediverse. For technical reasons only. CPU goes through the roof when new posts are boosted for the first time, causing 508 errors.
Does anyone else suffer from this problem and have any suggestions to lighten the load?
Maybe I’m overreacting and should just live with the minute or two of daily downtime?
I have Litespeed cache installed.
Affected sites: @dinotoyblog @blog@animaltoyforum.com @blog@monstertoyblog.com
@AdamStuartSmith @dinotoyblog @blog@animaltoyforum.com @blog@monstertoyblog.com so you have the issue even with the lightspeed cache?
-
@AdamStuartSmith @dinotoyblog @blog@animaltoyforum.com @blog@monstertoyblog.com so you have the issue even with the lightspeed cache?
@pfefferle Exactly. Maybe there are there some Litespeed settings I could tweak?
Thanks for your interest and help.
-
I thinking of uninstalling the excellent #Wordpress #ActivityPub Plugin and defederating my sites from the #Fediverse. For technical reasons only. CPU goes through the roof when new posts are boosted for the first time, causing 508 errors.
Does anyone else suffer from this problem and have any suggestions to lighten the load?
Maybe I’m overreacting and should just live with the minute or two of daily downtime?
I have Litespeed cache installed.
Affected sites: @dinotoyblog @blog@animaltoyforum.com @blog@monstertoyblog.com
@AdamStuartSmith@sauropods.win I had that problem with vanilla Mastodon for a bit when I first installed it. Any posts that got any sort of activity from anyone with a significant following would overwhelm my server for several minutes.
I solved that by setting up a #Varnish cache in front of my server. I don't use Mastodon anymore, but I still use the cache in front of my bespoke software (#Enigmatick). It works really well to take the load off of the core server for those fan-out posts. Varnish handles the load with aplomb.
-
@AdamStuartSmith@sauropods.win I had that problem with vanilla Mastodon for a bit when I first installed it. Any posts that got any sort of activity from anyone with a significant following would overwhelm my server for several minutes.
I solved that by setting up a #Varnish cache in front of my server. I don't use Mastodon anymore, but I still use the cache in front of my bespoke software (#Enigmatick). It works really well to take the load off of the core server for those fan-out posts. Varnish handles the load with aplomb.
@jdt @AdamStuartSmith Yep, this is the way. I recently did something similar with nginx https://join.piefed.social/2025/11/22/reducing-piefeds-dependence-on-cloudflare/
-
@jdt @AdamStuartSmith Yep, this is the way. I recently did something similar with nginx https://join.piefed.social/2025/11/22/reducing-piefeds-dependence-on-cloudflare/
@rimu@mastodon.nzoss.nz Definitely. Offloading the static assets to
nginxis a big win. Varnish adds a layer of serving from memory that takes it up a notch. Like having your own Fastly pop.It does require some configuration nuance to be sure you aren't serving cached assets to the wrong connections (e.g., authenticated
GETrequests that shouldn't be shared beyond a specific session). -
@pfefferle Exactly. Maybe there are there some Litespeed settings I could tweak?
Thanks for your interest and help.
@AdamStuartSmith is this "only" a problem when you post new things?
so there is always the option to tweak the scheduling, to not send all pings at the same time. the only downside is, that this can take some more time until every of your follower receives the new post.
-
@AdamStuartSmith is this "only" a problem when you post new things?
so there is always the option to tweak the scheduling, to not send all pings at the same time. the only downside is, that this can take some more time until every of your follower receives the new post.
@AdamStuartSmith I will put it on the backlog.
idea: an option to chose between different scaling/load levels.
1. publish as fast as possible (high load)
2. reduce bulks and increase pauses between sending
3. take your time
️ -
@AdamStuartSmith I will put it on the backlog.
idea: an option to chose between different scaling/load levels.
1. publish as fast as possible (high load)
2. reduce bulks and increase pauses between sending
3. take your time️ -
@AdamStuartSmith is this "only" a problem when you post new things?
so there is always the option to tweak the scheduling, to not send all pings at the same time. the only downside is, that this can take some more time until every of your follower receives the new post.
@pfefferle Only when I post new things, but for each new post there are two CPU peaks associated with CPU faults. One when it is first published, then a larger one later when I (or someone else with a moderate following) boosts it directly on Mastodon.
Happy for pings to take their time. Thanks for looking into it, I’m also working through the suggestions as far as I’m able.
-
@pfefferle Only when I post new things, but for each new post there are two CPU peaks associated with CPU faults. One when it is first published, then a larger one later when I (or someone else with a moderate following) boosts it directly on Mastodon.
Happy for pings to take their time. Thanks for looking into it, I’m also working through the suggestions as far as I’m able.
@AdamStuartSmith phew... the load as a result of boosting is hard to intercept: https://ar.al/2022/11/09/is-the-fediverse-about-to-get-fryed-or-why-every-toot-is-also-a-potential-denial-of-service-attack/
