Post B8ZL0xGcvaNblgRmyG by Oleksii@social.linux.pizza
 (DIR) More posts by Oleksii@social.linux.pizza
 (DIR) Post #B8ZHY7Cqb0qUKiEgng by selea@social.linux.pizza
       0 likes, 0 repeats
       
       I've had issues with Linux.Pizza's Matrix-server, mainly due to uncontrolled growth of the database. Even thou retention is set do one month, there is still an absurd amount of growth in the database.I have never encountered this before, not even with thousands of active users.I do know that other users had this issue before and have been vocal about it.Is there any solution to this yet?#matrix #linux #postgres
       
 (DIR) Post #B8ZIGWiF7okABJ68Rs by Frisk@woof.tech
       0 likes, 0 repeats
       
       @selea Do you do full vacuum of the database from time to time? Not a Postgres expert, but while Synapse does redact records, they still need to be cleaned up from the database to reclaim space.
       
 (DIR) Post #B8ZJwHO8ag2MlqAPjc by fraca7@social.linux.pizza
       0 likes, 0 repeats
       
       @selea DROP DATABASE MatrixI’ll see myself out
       
 (DIR) Post #B8ZKnN32HLCrfhc2t6 by selea@social.linux.pizza
       0 likes, 0 repeats
       
       @Frisk I have not, it will bring the database offline thou - maybe worth it
       
 (DIR) Post #B8ZL0xGcvaNblgRmyG by Oleksii@social.linux.pizza
       0 likes, 0 repeats
       
       @selea i dont know details, but IMO,  xmpp has same functionality,  but less resource intensive. Matrix is good to keep history of communication
       
 (DIR) Post #B8ae0nn2wxTASagzNA by selea@social.linux.pizza
       0 likes, 0 repeats
       
       @kitkat @Frisk what is "project hydra"?
       
 (DIR) Post #B8alHKAv0hGpDGgSW0 by Frisk@woof.tech
       0 likes, 0 repeats
       
       @selea @kitkat Ongoing project in Matrix focused on improving security of federation. I think what kitkat refers to, is the mechanism currently being worked on, which will allow servers to not concern themselves with thousands of state events (which are the main source of database space utilization and are unaffected by retention policies) as every room will have "milestone" events that will present state of the room at given time without the need of handling the entire chain until that milestone event. Effectively allowing a server to only care about room events only from last milestone event to now.Currently servers have to download and store entire state chains which in big rooms are huge.At least I think it went like that, truth be told, I've already forgot where I've read about it :blobcatsweats:
       
 (DIR) Post #B8ceJEObsVAMbj3dmi by selea@social.linux.pizza
       0 likes, 0 repeats
       
       @Frisk Thank you for the explanation :)
       
 (DIR) Post #B8ceUYSRd3Lk0uFcDw by selea@social.linux.pizza
       0 likes, 0 repeats
       
       I had to bring the instance down, and get a new drive in order to be able to shrink the DB.
       
 (DIR) Post #B8cfHBmuhgBPtU12DQ by selea@social.linux.pizza
       0 likes, 0 repeats
       
       I am currently really happy that I can use @joacim backup NAS as my temporary storage in order to fix this issue :D
       
 (DIR) Post #B8cfmV7wqnJiqnmjyq by joacim@mastodon.melin.org
       0 likes, 0 repeats
       
       @selea Mi casa su casa
       
 (DIR) Post #B8dI7biHM6zlcGGn56 by selea@social.linux.pizza
       0 likes, 0 repeats
       
       Database is shrunk to 180gb instead of 600gb - that's ok
       
 (DIR) Post #B8dIDOiMb5NH2U4QYS by selea@social.linux.pizza
       0 likes, 0 repeats
       
       Moving it back into place and will start the instance again, ETA 30 minutes
       
 (DIR) Post #B8dNKOVUGd8aC1Cf8y by c_merriweather@social.linux.pizza
       0 likes, 0 repeats
       
       @selea Thank you.💐
       
 (DIR) Post #B8dTw9ohvqibMUaEhU by selea@social.linux.pizza
       0 likes, 0 repeats
       
       @c_merriweather Sorry, it took longer for some reason.. It is back up since 20:00 CET - and with 500% more RAM and CPU. Hopefully that compensates for the downtime..