00:45:59 r​edsh4de:matrix.org: PR for partially solving #474 (no ETA): https://github.com/Cuprate/cuprate/pull/588
13:10:08 b​oog900:monero.social: I tested a sync on a PI 4 2 GB with a USB connected HDD. The most potato setup I could think of.
13:10:09 b​oog900:monero.social: Although I had to tune down the config values it did manage to get through the fast sync blocks in 16 hrs 10 minutes. After that it kept going for 5 hrs doing full verification before it OOMed, I have stopped it.
13:10:11 b​oog900:monero.social: I am happy with this as this would have probably taken months to fast sync on this setup with the old DB.
13:52:34 r​edsh4de:matrix.org: blazingly fast meme proven right once again
15:14:43 sech1: The most potato setup is SD card instead of HDD
15:25:58 b​oog900:monero.social: I don't have one big enough :)
15:33:50 r​edsh4de:matrix.org: i do have one of those 400GB SD cards somewhere... will give it a try if i find it
15:34:03 r​edsh4de:matrix.org: on a microPC with a Celeron N4100
15:51:45 s​yntheticbird:monero.social: I have a 2TB microsd card
15:52:16 s​yntheticbird:monero.social: but i don't have a rbpi
17:13:53 plowsof: Sell the card and buy a pi
17:50:43 b​oog900:monero.social: Meeting here in 10 minutes
18:00:21 b​oog900:monero.social: !meeting
18:00:23 m​oo:monero.social: Starting meeting: https://github.com/monero-project/meta/issues/1342
18:00:26 b​oog900:monero.social: 1) greetings
18:01:12 h​into:monero.social: hello
18:03:45 b​oog900:monero.social: 2) updates
18:05:53 b​oog900:monero.social: Me: testing the new db and fixing the issues I ran into
18:06:26 b​oog900:monero.social: A txpool issue was particularly annoying this debug
18:06:40 b​oog900:monero.social: To*
18:08:09 h​into:monero.social: me: took/taking a break from Cuprate for now and have been working on another related project
18:10:26 b​oog900:monero.social: 3) Project: What is next for Cuprate?
18:10:48 b​oog900:monero.social: Are you able to reveal the related project :)
18:12:34 h​into:monero.social: perhaps, eventually, maybe
18:13:09 h​into:monero.social: I can say that it uses Cuprate and monero-oxide :)
18:14:31 h​into:monero.social: was this OOM behavior happening with the old DB? IIRC I had the current cuprated running on a Pi 5 with a 2GB systemd limit
18:15:51 b​oog900:monero.social: No clue I don't think anyone was crazy enough to sync on a system like that
18:17:08 b​oog900:monero.social: The new db does use more ram though
18:18:28 h​into:monero.social: should Cuprate be prioritizing speed over resource footprint?
18:19:51 b​oog900:monero.social: Up to a point
18:21:55 h​into:monero.social: if we have a belief of who should be node operators in the future then we can focus on one of the qualities moreso
18:23:59 h​into:monero.social: if disk/memory prices keep spiking perhaps we'll all be priced out sooner than later :D
18:24:18 selsta: I have a question, afaik LMDB was always chosen for read speed and not the write speed, so we knew that initial sync would be slower in favour of faster reads later. How is that with the new DB system chosen by Cuprate?
18:25:49 b​oog900:monero.social: IMO 2 GB is already below the minimum amount of RAM for a monero node
18:26:29 b​oog900:monero.social: The fact we can fast sync and probably can full verify sync shows the DB doesn't use crazy amounts of RAM not that 2 GB is fine for cuprated
18:26:56 b​oog900:monero.social: Like itll only take a bit of spam to OOM a node with 2 GB on the old DB
18:34:19 b​oog900:monero.social: Selsta: the tapes DB is faster than LMDB for random reads. The tapes db is used for RCT outputs so for that we should be more efficient. For pre rct it would be slower however in my tests this doesn't actually slow it down at all. Presumably lookups make up a much smaller proportion of the time compared to the other things that need to be done during verification.
18:35:35 b​oog900:monero.social: I did a full verification sync in under 15 hrs with the new DB
18:36:39 selsta: really impressive and shows that there are still a lot of possible performance gains also in monerod
18:40:59 b​oog900:monero.social: Anything else to discuss today?
18:43:14 b​oog900:monero.social: Ah I don't think I answered why fjall was chosen
18:44:17 b​oog900:monero.social: It was chosen as its underlying "format" is a LSM, which is much better at writes than LMDB. Plus its in rust
18:44:31 b​oog900:monero.social: We could have gone with rocksDB
18:44:35 s​yntheticbird:monero.social: no
18:44:39 s​yntheticbird:monero.social: rocksDb is dogshit
18:44:47 b​oog900:monero.social: Lol
18:45:02 b​oog900:monero.social: Yeah its also doesn't have a good reputation
18:48:13 b​oog900:monero.social: I'll end here
18:48:24 b​oog900:monero.social: Thanks everyone
18:48:34 b​oog900:monero.social: !meeting
18:48:41 m​oo:monero.social: - Logs: https://github.com/monero-project/meta/issues/1342#issuecomment-3954040271
18:48:41 m​oo:monero.social: - Next meeting: https://github.com/monero-project/meta/issues/1345
18:51:07 b​oog900:monero.social: It was chosen as its underlying "format" is a LSM tree, which is much better at writes than LMDB. Plus its in rust