Conversation

Notices

1. FWC (fwc@0.5dollah.click)'s status on Friday, 03-Jan-2025 22:55:28 JST FWC

   • Griffith
   • bigtony
   • mischievoustomato
   • Jelqer Skelter

   Um, so the Mitra media folder is already 139 Gb. Who is responsible for this? @griffith @bigtony @mischievoustomato @sneeden

   • mischievoustomato@0.5dollah.click's status on Friday, 03-Jan-2025 22:55:25 JST mischievoustomato

     in reply to

     • Griffith
     • bigtony
     • Jelqer Skelter

     @griffith @fwc @bigtony @sneeden yeees

   • mischievoustomato@0.5dollah.click's status on Friday, 03-Jan-2025 22:55:25 JST mischievoustomato

     in reply to

     • Griffith
     • bigtony
     • Jelqer Skelter

     @fwc @bigtony @griffith @sneeden i just ran mitractl delete-orphaned-files and uh... it's stuck there

   • Griffith (griffith@0.5dollah.click)'s status on Friday, 03-Jan-2025 22:55:26 JST Griffith

     in reply to

     • bigtony
     • mischievoustomato
     • Jelqer Skelter

     @mischievoustomato @fwc @bigtony @sneeden I’ve posted less on here than on our Pleroma instance, so it shouldn’t be that big.

   • mischievoustomato@0.5dollah.click's status on Friday, 03-Jan-2025 22:55:27 JST mischievoustomato

     in reply to

     • Griffith
     • bigtony
     • Jelqer Skelter

     @fwc @bigtony @griffith @sneeden all of us essentially, but I don't remember mine being that big, let me check

   • FWC (fwc@0.5dollah.click)'s status on Friday, 03-Jan-2025 22:55:54 JST FWC

     in reply to

     • Griffith
     • bigtony
     • mischievoustomato
     • Jelqer Skelter

     @mischievoustomato @bigtony @griffith @sneeden This has to be a bug of some kind, no?

   • silverpill (silverpill@mitra.social)'s status on Sunday, 05-Jan-2025 03:59:31 JST silverpill

     in reply to

     • Griffith
     • bigtony
     • 𝔧𝔞𝔢
     • mischievoustomato
     • Jelqer Skelter

     @jae @griffith @fwc @mischievoustomato @sneeden @bigtony

     When data retention parameters are changed, mitra starts removing unnecessary posts from the database and associated media from the filesystem. Media deletion happens in background tasks which block other background tasks, including incoming federation (not outgoing, which uses different thread).

     If the number of deleted posts is huge, processing of media cleanup tasks may take a long time, leading to a noticeable delay in federation. But this doesn't lead to data loss.

   • 𝔧𝔞𝔢 (jae@wizard.casa)'s status on Sunday, 05-Jan-2025 03:59:32 JST 𝔧𝔞𝔢

     in reply to

     • silverpill
     • Griffith
     • bigtony
     • mischievoustomato
     • Jelqer Skelter

     @fwc @silverpill @griffith @bigtony @mischievoustomato @sneeden

     > Postgres will use 100% CPU to delete remote posts & media until it's done after a few hours.

     i'm going to try it with an instance i'm synthetically filling up to see if i can offer some insight. my guess is that it's based on the amount of media objects/etc which determines how long it will exhaust the threads. there may be a way to optimize the query statement or general operation

   • 𝔧𝔞𝔢 (jae@wizard.casa)'s status on Sunday, 05-Jan-2025 03:59:33 JST 𝔧𝔞𝔢

     in reply to

     • Griffith
     • bigtony
     • mischievoustomato
     • Jelqer Skelter

     @mischievoustomato @fwc @griffith @bigtony @sneeden

     > i'm not sure. postgres is using 100% cpu, i assume it's reading shit, but at the same time my instance is in a non working state due to well, expired domain xDDD

     it's unlikely postgres is banging all the cores at 100% due to an expired domain. start dumping out some postgres logs to see what's happening. this should not be happening unless you're on some single core vps and you've got some other process hitting it hard.

   • FWC (fwc@0.5dollah.click)'s status on Sunday, 05-Jan-2025 03:59:33 JST FWC

     in reply to

     • silverpill
     • Griffith
     • bigtony
     • 𝔧𝔞𝔢
     • mischievoustomato
     • Jelqer Skelter

     @jae @bigtony @griffith @mischievoustomato @sneeden

     No, this is what happened because he changed (reduced) the remote post retention. Same thing happened to 0.5dollah.click yesterday after I changed ours. Postgres will use 100% CPU to delete remote posts & media until it's done after a few hours. Until that happens, there's no incoming federation. I let @silverpill know about it.

   • mischievoustomato@0.5dollah.click's status on Sunday, 05-Jan-2025 03:59:34 JST mischievoustomato

     in reply to

     • Griffith
     • bigtony
     • Jelqer Skelter

     @fwc @bigtony @griffith @sneeden i'm not sure. postgres is using 100% cpu, i assume it's reading shit, but at the same time my instance is in a non working state due to well, expired domain xDDD

   • 𝔧𝔞𝔢 (jae@wizard.casa)'s status on Monday, 06-Jan-2025 11:06:46 JST 𝔧𝔞𝔢

     in reply to

     • silverpill
     • Griffith
     • bigtony
     • mischievoustomato
     • Jelqer Skelter

     @fwc @silverpill @griffith @bigtony @mischievoustomato @sneeden agreed, a note would work.

   • silverpill (silverpill@mitra.social)'s status on Monday, 06-Jan-2025 11:06:46 JST silverpill

     in reply to

     • Griffith
     • bigtony
     • 𝔧𝔞𝔢
     • mischievoustomato
     • Jelqer Skelter

     @jae @griffith @fwc @mischievoustomato @sneeden @bigtony This shouldn't be very difficult to fix, I created a tracking issue https://codeberg.org/silverpill/mitra/issues/86

   • FWC (fwc@0.5dollah.click)'s status on Monday, 06-Jan-2025 11:06:47 JST FWC

     in reply to

     • silverpill
     • Griffith
     • bigtony
     • 𝔧𝔞𝔢
     • mischievoustomato
     • Jelqer Skelter

     @silverpill @bigtony @griffith @jae @mischievoustomato @sneeden

     Yep, it's probably just something to note in the config.yaml or when the default changes in Mitra v4.0.0