-
h-erectus
is there any hope for a decentralised cryptocurrency that's scalable?
-
h-erectus
is there any theorem that proves that this is impossible?
-
h-erectus
i think cryptocurrency is essentially: cryptographically proving that a chunk of data came out of 'genesis block', and a way to detect double spending.
-
gingeropolous
selsta, so im thinking my issues are because systemd can't handle LMDBs memory magic
-
m-relay
<ofrnxmr:monero.social> It works without systemd? (Doesnt sound right)
-
selsta
gingeropolous: does it happen on master or release nodes too?
-
plowsof
Custom systemd script?
-
gingeropolous
tesing now on release 18.3.1. It takes a day or so to get up to 40GB.
-
hyc
what's happening? I can't find the beginning of that convo. systemd complains at 40GB?
-
hyc
you can force monerod to cycle thru a lot of RAM, just try to refresh a wallet from block 1
-
m-relay
<orion_midast:matrix.org> Hey is there any chance we can try to get on Tangem hardware wallet? Its fully open source
-
m-relay
<ofrnxmr:monero.social> It uses 40gb of ram? (Cant we limit the sync so it doesnt download so far ahead?)
-
m-relay
<ofrnxmr:monero.social> Ask them
-
hyc
ofrnxmr: it doesn't need 40GB of RAM. it uses as much as is free.
-
m-relay
<orion_midast:matrix.org> I did in their discord sever. They said they added it to the list but thats some retard stuff.
-
m-relay
<orion_midast:matrix.org> do we have dev power to help them?
-
m-relay
<orion_midast:matrix.org> or something
-
m-relay
<ofrnxmr:monero.social> Hyc - can it be limited with a flag?
-
m-relay
<ofrnxmr:monero.social> Something like `block-download-max-size` but for total sync cache?
-
hyc
there is no sync cache
-
hyc
LMDB only uses free RAM. If LMDB is using it, that means nothing else on the system wanted it.
-
hyc
If anything else on the system wants RAM, the RAM LMDB was using will get taken over by the other app, instantly and transparently.
-
m-relay
<trasherdk:monero.social> I only had a few (100KB -> 1MB) swap activated on a wallet sync. 2 threads, 2 G RAM.
-
m-relay
<trasherdk:monero.social> Eh, 4 threads and 4 G RAM, 360 M swap.
-
m-relay
<trasherdk:monero.social> Daemon:
-
m-relay
<trasherdk:monero.social> ```2023-12-18 15:53:48.272 I Found block <c4bbfb2df77f5467c6cecdcb9e5504e984455ad99448bdf2f5dbaca4ab0807c3> at height 1503370 for difficulty: 219498
-
m-relay
<trasherdk:monero.social> 2023-12-18 16:00:35.457 I Found block <477b174be12b3c66e6b844a2ec552c9d1882fa4974c758111860775522713e0c> at height 1503371 for difficulty: 218130
-
m-relay
<trasherdk:monero.social> 2023-12-18 16:13:45.678 I Found block <7f18ef7fbdea0daf137fb98bf92d07e2f4cf77e208a9291d3fdd4484d14a7890> at height 1503378 for difficulty: 217964
-
m-relay
<trasherdk:monero.social> ```
-
m-relay
<trasherdk:monero.social> Wallet:
-
m-relay
<trasherdk:monero.social> ```
-
m-relay
<trasherdk:monero.social> 2023-12-18 15:51:08.084 W Received money: 2.063935154664, with tx: <1fec7f486687d3f96c15786285295232a0be17adaf43b4aaf9095cdc89b7185e>
-
m-relay
<trasherdk:monero.social> 2023-12-18 15:53:48.320 W Received money: 2.063778998310, with tx: <dfe0dd9a6bc760b2bb1dc0e438a886f8bf84159a34ce8ddf5a0939003a432d32>
-
m-relay
<trasherdk:monero.social> 2023-12-18 15:53:48.320 W Received money: 2.063775061964, with tx: <823307248b284b862e7d1a4424af6a5d70248f5e33b858515b2ffe0883cd922e>
-
m-relay
<trasherdk:monero.social> 2023-12-18 16:00:49.169 W Received money: 2.063771125626, with tx: <c7d19a261b38ef7bce59aa775eaed6dc57f92efc044ec3efbad995beee4c5cf9>
-
m-relay
<trasherdk:monero.social> 2023-12-18 16:13:50.381 W Received money: 2.068908371466, with tx: <c8718e6cf658163cdb117b42e4c4a28d20f7eed57d7289d648614f783e7521a6>
-
m-relay
<trasherdk:monero.social> ```
-
hyc
pastebin, dude
-
m-relay
<trasherdk:monero.social> The short version is, daemon mining and wallet syncing, no more than 50% of 4G RAM is used, and at some point, 360M swap was active.
-
hyc
You should try setting /proc/sys/vm/swappiness = 0. there's no reason for swap to get used if you have free RAM.
-
hyc
a wallet that is up to date won't cause much memory use because it's only going to sync against newly received blocks. it won't cause any more memory to be used than monerod itself would have used anyway.
-
hyc
mining will use 2.3GB
-
m-relay
<trasherdk:monero.social> `2.73G/3.82G` to be precise 😎
-
m-relay
<trasherdk:monero.social> The swap usage could be for a number of other reasons. MySQ, Apache and kiddies hitting the firewall.
-
hyc
syncing a wallet from far in the past will force monerod to read older blocks that it wasn't referencing up till then, so RAM use should go up
-
hyc
I just did a rescan_bc soft 1000000 and I can see the RAM use go up on my monerod now ...
-
hyc
from 1000000 to 1851000 added 3.5GB to the process size
-
hyc
rather, to the Shared memory use
-
m-relay
<trasherdk:monero.social> I'll try to launch a daemon sync and see how that affect memory. I think mainnet is about 2-4K blocks behind on this box.
-
m-relay
<trasherdk:monero.social> Well, it does snatch up a bit, but not much.
-
m-relay
-
m-relay
-
m-relay
<trasherdk:monero.social> Does images cross the bridges?
-
moneromooo
You posted two links to matrix.monero.social before that question. They could be images.
-
m-relay
<trasherdk:monero.social> So I guess they don't cross.
-
rbrunner
Well, right now the links do work. But maybe not everybody is in the mood to click :)
-
m-relay
<trasherdk:monero.social> LOL. Okay. Launching a second daemon added aprox. 2G to memory usage.
-
m-relay
<trasherdk:monero.social> A few. 20-30M swap.
-
gingeropolous
-
gingeropolous
basically, at some point in time (for a reason I forgot), i instructed systemd to do some memory management for monerod
-
gingeropolous
when I pulled in that weak_ptrs PR 7345, i noticed that monerod was being restarted by systemd because it was hitting my 40GB limit
-
gingeropolous
so now im trying to see if release has the same behavior when managed by systemd
-
gingeropolous
sorry for jumping back and forth with communication channels. im never sure which of my musings warrant the permanent record in the githubs
-
hyc
ah. if your box has 64GB RAM then sure, monerod may try to use more than 40GB. is this a public node?
-
hyc
again, if it's only syncing recent blocks, memory use would never get that high on its own
-
plowsof
If its the same box as xmrchain, is it possible that its the onion blockchain explorer? People using the api? (Ive a 64gb box that is using between 5-10gb)
-
hyc
yeah, if they're all looking up random stuff, sure
-
hyc
anyway the thing to watch out for is whether the memory use is mostly shared memory or not. if it's mostly shared, that's LMDB and filesystem cache, and nothing to worry about.
-
hyc
if shared is not the majority, then there may be a monerod memory leak.
-
sech1
"If anything else on the system wants RAM, the RAM LMDB was using will get taken over by the other app, instantly and transparently" But what if 2 LMDB instances are running together, like monerod and xmrblocks (the block explorer)?
-
sech1
Right now I have 49.3 GiB for monerod and 7.5 GiB for xmrblocks on p2pool.io
-
sech1
On a 64 GiB system
-
Inge
hm my monerod is only using 1.6GB of RAM and the system has good deal more available
-
m-relay
<ofrnxmr:monero.social> Try remote syncing an old wallet (maybe)
-
hyc
sech1: it's all handled by the kernel filesystem cache. Typically some advanced form of LRU cache replacement.