Conversation

Notices

1. Embed this notice
   vib [:blobcatblush:] (vib@fedi.absturztau.be)'s status on Tuesday, 25-Oct-2022 04:02:00 JST vib [:blobcatblush:]

   • Puniko ?
   • :blobcathug:

   @Jain @puniko I will probably just let it finish but yes all of my CPU cores are close to 100% and now new posts won’t federate until the queue gets emptier

   • Embed this notice
     Puniko ? (puniko@mk.absturztau.be)'s status on Tuesday, 25-Oct-2022 04:01:59 JST Puniko ?

     in reply to

     • :blobcathug:

     @vib@fedi.absturztau.be @Jain@blob.cat tasks usually should hang around for 2 days or so, if they continue to fail

   • Embed this notice
     Puniko ? (puniko@mk.absturztau.be)'s status on Tuesday, 25-Oct-2022 04:02:31 JST Puniko ?

     in reply to

     • :blobcathug:

     @vib@fedi.absturztau.be @Jain@blob.cat you could also clear the queue as a last resort btw

   • Embed this notice
     Tosti ? (tost@mk.toast.cafe)'s status on Tuesday, 25-Oct-2022 23:44:08 JST Tosti ?

     in reply to

     • Puniko ?
     • :blobcathug:

     @puniko@mk.absturztau.be @vib@fedi.absturztau.be @Jain@blob.cat we have a new mechanism that instantly drops tasks as they come up if they target a presumed-dead (automatically detected) instance, so while the queue is huge it should also resolve VERY fast

     Puniko ? likes this.
   • Embed this notice
     Puniko ? (puniko@mk.absturztau.be)'s status on Tuesday, 25-Oct-2022 23:47:34 JST Puniko ?

     in reply to

     • Tosti ?
     • :blobcathug:

     @tost@mk.toast.cafe @Jain@blob.cat @vib@fedi.absturztau.be ah, good to know that i dont have to worry about dead instances anymore

   • Embed this notice
     Tosti ? (tost@mk.toast.cafe)'s status on Tuesday, 25-Oct-2022 23:47:40 JST Tosti ?

     in reply to

     • Puniko ?
     • :blobcathug:

     @Jain@blob.cat @vib@fedi.absturztau.be @puniko@mk.absturztau.be
     not custom, it's in main
     this is the primary commit: https://akkoma.dev/FoundKeyGang/FoundKey/commit/91a4f38871e4f39fc30e943fa913dd9ab43cc600 (minus cleanup)
     the tl;dr is that we already keep a statistic for "the last time we successfully communicated with an instance" (receiving an activity, sending an activity without an error, etc)
     if the last time that has happened is > a month, the instance is considered "dead for now" and we stop sending activities (exactly the same way as if it was suspended via the "stop sending activities" button)
     all that it takes to undo that is for us to receive an activity from them
     iirc a fetch (e.g via search or querying a user) should be sufficient, or to just have them send us something
     i.e truly automated "dead instance detection"

     Puniko ? likes this.
   • Embed this notice
     :blobcathug: (jain@blob.cat)'s status on Tuesday, 25-Oct-2022 23:47:41 JST :blobcathug:

     in reply to

     • Puniko ?
     • Tosti ?
     @tost @vib @puniko custom patch? are instances which are temporarely down affected by this?