bitcoin/doc/reduce-memory.md
Tim Ruffing 12f7257b8f
doc: Be vague instead of wrong about MALLOC_ARENA_MAX
Before this commit, we claim that glibc's malloc implementation uses 2
arenas by default. But that's true only on 32-bit systems, and even
there, it uses *up* to 2 arenas.

This commit fixes the wrong statement. The new statement is
intentionally vague to reduce our maintenance burden.

For details, see:
https://www.gnu.org/software/libc/manual/html_node/Memory-Allocation-Tunables.html#index-glibc_002emalloc_002earena_005fmax

Noticed in:
https://github.com/bitcoin/bitcoin/pull/27642#issuecomment-1728103427
2023-09-20 17:12:24 +00:00

3.5 KiB

Reduce Memory

There are a few parameters that can be dialed down to reduce the memory usage of bitcoind. This can be useful on embedded systems or small VPSes.

In-memory caches

The size of some in-memory caches can be reduced. As caches trade off memory usage for performance, reducing these will usually have a negative effect on performance.

  • -dbcache=<n> - the UTXO database cache size, this defaults to 450. The unit is MiB (1024).
    • The minimum value for -dbcache is 4.
    • A lower -dbcache makes initial sync time much longer. After the initial sync, the effect is less pronounced for most use-cases, unless fast validation of blocks is important, such as for mining.

Memory pool

  • In Bitcoin Core there is a memory pool limiter which can be configured with -maxmempool=<n>, where <n> is the size in MB (1000). The default value is 300.

    • The minimum value for -maxmempool is 5.
    • A lower maximum mempool size means that transactions will be evicted sooner. This will affect any uses of bitcoind that process unconfirmed transactions.
  • Since 0.14.0, unused memory allocated to the mempool (default: 300MB) is shared with the UTXO cache, so when trying to reduce memory usage you should limit the mempool, with the -maxmempool command line argument.

  • To disable most of the mempool functionality there is the -blocksonly option. This will reduce the default memory usage to 5MB and make the client opt out of receiving (and thus relaying) transactions, except from peers who have the relay permission set (e.g. whitelisted peers), and as part of blocks.

    • Do not use this when using the client to broadcast transactions as any transaction sent will stick out like a sore thumb, affecting privacy. When used with the wallet it should be combined with -walletbroadcast=0 and -spendzeroconfchange=0. Another mechanism for broadcasting outgoing transactions (if any) should be used.

Number of peers

  • -maxconnections=<n> - the maximum number of connections, which defaults to 125. Each active connection takes up some memory. This option applies only if inbound connections are enabled; otherwise, the number of connections will not be more than 11. Of the 11 outbound peers, there can be 8 full-relay connections, 2 block-relay-only ones, and occasionally 1 short-lived feeler or extra outbound block-relay-only connection.

  • These limits do not apply to connections added manually with the -addnode configuration option or the addnode RPC, which have a separate limit of 8 connections.

Thread configuration

For each thread a thread stack needs to be allocated. By default on Linux, threads take up 8MiB for the thread stack on a 64-bit system, and 4MiB in a 32-bit system.

  • -par=<n> - the number of script verification threads, defaults to the number of cores in the system minus one.
  • -rpcthreads=<n> - the number of threads used for processing RPC requests, defaults to 4.

Linux specific

By default, glibc's implementation of malloc may use more than one arena. This is known to cause excessive memory usage in some scenarios. To avoid this, make a script that sets MALLOC_ARENA_MAX before starting bitcoind:

#!/usr/bin/env bash
export MALLOC_ARENA_MAX=1
bitcoind

The behavior was introduced to increase CPU locality of allocated memory and performance with concurrent allocation, so this setting could in theory reduce performance. However, in Bitcoin Core very little parallel allocation happens, so the impact is expected to be small or absent.