encountering an OOM problem about bsc HOT 4 CLOSED

based the your pprof file, seems like the knownTxs.Add() occupies lots of memory, was your node connected to lots of peers with large pending transaction needs to be broadcasted.
But the pprof is quite low level, while Golang would have GC to release the temporary memory of knownTxs.Add(), so this pprof may not that accurate.

from bsc.

thank you , i find the config.yaml

the max peer maybe too large 5000
amd our geth connect too many peer

i will try change MaxPeers less than 100

from bsc.

Try not specify the body cache and batch size: --bodies.cache=214748364800 --batchSize=4096M.
These 2 flags will bring heavy to RAM caches. Similar issue: node-real/bsc-erigon#39

from bsc.

System information

Geth version: geth v1.4.6 OS & Version: AWS Amazon Linux x86, with 32GB of memory

Our machine running the BSC full node gets its BSC process killed after running for several days. Checking the system message logs shows an OOM error. Below is the memory profile graph we observed using pprof.

The memory was still increasing when using pprof. Could you please let me know the typical maximum memory usage of a BSC full node? Is it possible that the machine does not have enough memory?

I think BSC full node will generally take ~20GB. Could u provide your config.toml and cmd line? We can check if there is any unexpected memory usage.

from bsc.