Ethereum implementation on the efficiency frontier
You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
Go to file
Andrew Ashikhmin d8d16c3c7c
Fix Errorf (#8241)
Small fix after PR #8239
5 days ago
.buildkite Rename to Erigon (#2018) 2 years ago
.github Downloader: add tests to protect from re-downloading (#8191) 1 week ago
accounts/abi Update execution-spec-tests to v1.0.1 (#7980) 2 months ago
cl Fixed forkchoice nil exception (#8223) 7 days ago
cmd Milestone stage processing (#8187) 6 days ago
common Support for --chain=holesky (#8064) 4 weeks ago
consensus Fix Errorf (#8241) 5 days ago
core Implement EIP-7516: BLOBBASEFEE opcode (#8231) 5 days ago
crypto Replace --override.shanghaiTime flag with --override.cancun (#7964) 2 months ago
dataflow Header downloader monitor for diagnostics system (#7590) 4 months ago
design auto-format code by prettier (similar to gofmt) (#405) 4 years ago
diagnostics Heimdall metrics + Metrics HTTP server rationalization (#8094) 3 weeks ago
docs rename torquem-ch project 1 month ago
eth Add flag test to prevent milestone services from starting (#8239) 5 days ago
ethdb Fixed pending block response (#8200) 1 week ago
ethstats blockReader: use in ethstat (#7565) 4 months ago
event rename (#1978) 2 years ago
hive feat(hive): update exclusions after timeouts fix (#5490) 1 year ago
hooks docker_hub_default_pid (#4819) 1 year ago
k8s 7667 cli params vhost (#7669) 4 months ago
metrics Group metrics by type and fix missed label stripping (#8186) 1 week ago
migrations Move ChainID from CommonTx to DynamicFeeTransaction (#7732) 3 months ago
node torrent: add --webseeds cli arg (#8176) 2 weeks ago
p2p p2p: limit ping requests from a single peer (#8113) 3 weeks ago
params holesky genesis update (#8225) 5 days ago
rlp p2p: refactor peer errors to propagate with a DiscReason (#8089) 3 weeks ago
rpc moved metrics sub packages types to metrics (#8119) 3 weeks ago
spectest Better caplin logging and logic (#7992) 1 month ago
tests Upgrade execution-spec-tests to Cancun Devnet 9 Release (#8208) 6 days ago
turbo Milestone stage processing (#8187) 6 days ago
visual Continue comparison of genesis block with geth, expand long values (#223) 4 years ago
.dockerignore make Docker build smarter, add Dockerfile.debian (#6344) 9 months ago
.env.example backward compatibility: use default UID=1000 GID=1000 (#4702) 1 year ago
.gitattributes .gitattributes: enable solidity highlighting on github (#16425) 6 years ago
.gitignore Milestone stage processing (#8187) 6 days ago
.gitmodules remove libmdbx git-submodule, make db-tools work on vendored to mdbx-go .c code (after "make dist") (#5392) 1 year ago
.golangci.yml Downloader: add tests to protect from re-downloading (#8191) 1 week ago
.goreleaser.yml remove netgo tag on win (#7421) 5 months ago
.readthedocs.yml first sphinx doc portion (#1144) 3 years ago
AUTHORS build: deduplicate same authors with different casing 4 years ago
COPYING COYPING: restore the full text text of GPL (#21568) 3 years ago
COPYING.LESSER all: update license information 8 years ago Block Gas limit contract integration for Chiado (#7727) 3 months ago
Dockerfile Prepare regression testings for kube (#7697) 4 months ago
Dockerfile.debian Forgot to configure flags (#7701) 4 months ago
Dockerfile.release chore: add multi-arch goreleaser workflow (#5185) 1 year ago
Makefile fix macos user homedir in makefile (#8133) 3 weeks ago Update readme ports (#8205) 1 week ago Fix: typos (#7710) 3 months ago Fix: typos (#7710) 3 months ago
debug.Dockerfile feat(docker): add debug.Dockerfile with delve (#7125) 6 months ago
docker-compose.yml deps up (after release) (#8142) 3 weeks ago
go.mod txpool: batch IdHashKnown calls - to reduce amount of mutex lock/unlock. less string(byte) casts (#8226) 6 days ago
go.sum txpool: batch IdHashKnown calls - to reduce amount of mutex lock/unlock. less string(byte) casts (#8226) 6 days ago
interfaces.go EIP-4844: rename "data hash" to "blob hash" (#7947) 2 months ago Nightly tests (#444) 3 years ago tests, secp256k1: secp256k1 fuzzer implementation (#2906) 2 years ago
rules.go Use go:build lines and remove obsolete +build lines (#4175) 1 year ago
tools.go move mdbx to new org (#8061) 1 month ago
wmake.ps1 [wip]: test non-nil compress.Next (#8072) 4 weeks ago


Erigon is an implementation of Ethereum (execution layer with embeddable consensus layer), on the efficiency frontier. Archive Node by default.

Build status


Disclaimer: this software is currently a tech preview. We will do our best to keep it stable and make no breaking changes but we don't guarantee anything. Things can and will break.

Important defaults: Erigon is an Archive Node by default (to remove history see: --prune flags in erigon --help). We don't allow change this flag after first start.

In-depth links are marked by the microscope sign (🔬)

System Requirements

  • For an Archive node of Ethereum Mainnet we recommend >=3TB storage space: 1.8TB state (as of March 2022), 200GB temp files (can symlink or mount folder <datadir>/temp to another disk). Ethereum Mainnet Full node ( see --prune* flags): 400Gb (April 2022).

  • Goerli Full node (see --prune* flags): 189GB on Beta, 114GB on Alpha (April 2022).

  • Gnosis Chain Archive: 370GB (January 2023).

  • Polygon Mainnet Archive: 5TB. Polygon Mumbai Archive: 1TB. (April 2022).

SSD or NVMe. Do not recommend HDD - on HDD Erigon will always stay N blocks behind chain tip, but not fall behind. Bear in mind that SSD performance deteriorates when close to capacity.

RAM: >=16GB, 64-bit architecture.

Golang version >= 1.19; GCC 10+ or Clang; On Linux: kernel > v4

🔬 more details on disk storage here and here.


Getting Started

For building the latest stable release (this will be suitable for most users just wanting to run a node):

git clone --branch stable --single-branch
cd erigon
make erigon

You can check the list of releases for release notes.

For building the bleeding edge development branch:

git clone --recurse-submodules
cd erigon
git checkout devel
make erigon

Default --snapshots for mainnet, goerli, gnosis, chiado. Other networks now have default --snapshots=false. Increase download speed by flag 🔬 See Downloader docs

Use --datadir to choose where to store data.

Use --chain=gnosis for Gnosis Chain, --chain=bor-mainnet for Polygon Mainnet, and --chain=mumbai for Polygon Mumbai. For Gnosis Chain you need a Consensus Layer client alongside Erigon (

Running make help will list and describe the convenience commands available in the Makefile.

Datadir structure

  • chaindata: recent blocks, state, recent state history. low-latency disk recommended.
  • snapshots: old blocks, old state history. can symlink/mount it to cheaper disk. mostly immutable. must have ~100gb free space (for merge recent files to bigger one).
  • temp: can grow to ~100gb, but usually empty. can symlink/mount it to cheaper disk.
  • txpool: pending transactions. safe to remove.
  • nodes: p2p peers. safe to remove.



  • verbosity
  • log.console.verbosity (overriding alias for verbosity)
  • log.json
  • log.console.json (alias for log.json)
  • log.dir.path
  • log.dir.prefix
  • log.dir.verbosity
  • log.dir.json

In order to log only to the stdout/stderr the --verbosity (or log.console.verbosity) flag can be used to supply an int value specifying the highest output log level:

  LvlCrit = 0
  LvlError = 1
  LvlWarn = 2
  LvlInfo = 3
  LvlDebug = 4
  LvlTrace = 5

To set an output dir for logs to be collected on disk, please set --log.dir.path If you want to change the filename prodiced from erigon you should also set the --log.dir.prefix flag to an alternate name. The flag --log.dir.verbosity is also available to control the verbosity of this logging, with the same int value as above, or the string value e.g. ' debug' or 'info'. Default verbosity is 'debug' (4), for disk logging.

Log format can be set to json by the use of the boolean flags log.json or log.console.json, or for the disk output --log.dir.json.


Erigon by default is "all in one binary" solution, but it's possible start TxPool as separated processes. Same true about: JSON RPC layer (RPCDaemon), p2p layer (Sentry), history download layer (Downloader), consensus. Don't start services as separated processes unless you have clear reason for it: resource limiting, scale, replace by your own implementation, security. How to start Erigon's services as separated processes, see in docker-compose.yml.

Embedded Consensus Layer

On Ethereum Mainnet, Görli, and Sepolia, the Engine API can be disabled in favour of the Erigon native Embedded Consensus Layer. If you want to use the internal Consensus Layer, run Erigon with flag --internalcl. Warning: Staking (block production) is not possible with the embedded CL.


If you would like to give Erigon a try, but do not have spare 2TB on your drive, a good option is to start syncing one of the public testnets, Görli. It syncs much quicker, and does not take so much disk space:

git clone --recurse-submodules -j8
cd erigon
make erigon
./build/bin/erigon --datadir=<your_datadir> --chain=goerli

Please note the --datadir option that allows you to store Erigon files in a non-default location, in this example, in goerli subdirectory of the current directory. Name of the directory --datadir does not have to match the name of the chain in --chain.

Block Production (PoW Miner or PoS Validator)

Disclaimer: Not supported/tested for Gnosis Chain and Polygon Network (In Progress)

Support only remote-miners.

  • To enable, add --mine --miner.etherbase=... or --mine --miner.miner.sigkey=... flags.
  • Other supported options: --miner.extradata, --miner.notify, --miner.gaslimit, --miner.gasprice , --miner.gastarget
  • JSON-RPC supports methods: eth_coinbase , eth_hashrate, eth_mining, eth_getWork, eth_submitWork, eth_submitHashrate
  • JSON-RPC supports websocket methods: newPendingTransaction
  • TODO:
    • we don't broadcast mined blocks to p2p-network yet, but it's easy to accomplish
    • eth_newPendingTransactionFilter
    • eth_newBlockFilter
    • eth_newFilter
    • websocket Logs

🔬 Detailed explanation is here.


Windows users may run erigon in 3 possible ways:

  • Build executable binaries natively for Windows using provided wmake.ps1 PowerShell script. Usage syntax is the same as make command so you have to run .\wmake.ps1 [-target] <targetname>. Example: .\wmake.ps1 erigon builds erigon executable. All binaries are placed in .\build\bin\ subfolder. There are some requirements for a successful native build on windows :

    • Git for Windows must be installed. If you're cloning this repository is very likely you already have it
    • GO Programming Language must be installed. Minimum required version is 1.19
    • GNU CC Compiler at least version 10 (is highly suggested that you install chocolatey package manager - see following point)
    • If you need to build MDBX tools (i.e. .\wmake.ps1 db-tools) then Chocolatey package manager for Windows must be installed. By Chocolatey you need to install the following components : cmake, make, mingw by choco install cmake make mingw. Make sure Windows System "Path" variable has: C:\ProgramData\chocolatey\lib\mingw\tools\install\mingw64\bin

    Important note about Anti-Viruses During MinGW's compiler detection phase some temporary executables are generated to test compiler capabilities. It's been reported some anti-virus programs detect those files as possibly infected by Win64/Kryptic.CIS trojan horse (or a variant of it). Although those are false positives we have no control over 100+ vendors of security products for Windows and their respective detection algorithms and we understand this might make your experience with Windows builds uncomfortable. To workaround the issue you might either set exclusions for your antivirus specifically for build\bin\mdbx\CMakeFiles sub-folder of the cloned repo or you can run erigon using the following other two options

  • Use Docker : see docker-compose.yml

  • Use WSL (Windows Subsystem for Linux) strictly on version 2. Under this option you can build Erigon just as you would on a regular Linux distribution. You can point your data also to any of the mounted Windows partitions ( eg. /mnt/c/[...], /mnt/d/[...] etc) but in such case be advised performance is impacted: this is due to the fact those mount points use DrvFS which is a network file system and, additionally, MDBX locks the db for exclusive access which implies only one process at a time can access data. This has consequences on the running of rpcdaemon which has to be configured as Remote DB even if it is executed on the very same computer. If instead your data is hosted on the native Linux filesystem non limitations apply. Please also note the default WSL2 environment has its own IP address which does not match the one of the network interface of Windows host: take this into account when configuring NAT for port 30303 on your router.

Using TOML or YAML Config Files

You can set Erigon flags through a YAML or TOML configuration file with the flag --config. The flags set in the configuration file can be overwritten by writing the flags directly on Erigon command line


./build/bin/erigon --config ./config.yaml --chain=goerli

Assuming we have chain : "mainnet" in our configuration file, by adding --chain=goerli allows the overwrite of the flag inside of the yaml configuration file and sets the chain to goerli


Example of setting up TOML config file

`datadir = 'your datadir'
port = 1111
chain = "mainnet"
http = true

"http.api" = ["eth","debug","net"]


Example of setting up a YAML config file

datadir : 'your datadir'
port : 1111
chain : "mainnet"
http : true
private.api.addr : "localhost:9090"

http.api : ["eth","debug","net"]

Beacon Chain (Consensus Layer)

Erigon can be used as an Execution Layer (EL) for Consensus Layer clients (CL). Default configuration is OK.

If your CL client is on a different device, add --authrpc.addr (Engine API listens on localhost by default) as well as --authrpc.vhosts <CL host> where <CL host> is your source host or any.

In order to establish a secure connection between the Consensus Layer and the Execution Layer, a JWT secret key is automatically generated.

The JWT secret key will be present in the datadir by default under the name of jwt.hex and its path can be specified with the flag --authrpc.jwtsecret.

This piece of info needs to be specified in the Consensus Layer as well in order to establish connection successfully. More information can be found here.

Once Erigon is running, you need to point your CL client to <erigon address>:8551, where <erigon address> is either localhost or the IP address of the device running Erigon, and also point to the JWT secret path created by Erigon.

Multiple Instances / One Machine

Define 6 flags to avoid conflicts: --datadir --port --http.port --authrpc.port --torrent.port --private.api.addr. Example of multiple chains on the same machine:

# mainnet
./build/bin/erigon --datadir="<your_mainnet_data_path>" --chain=mainnet --port=30303 --http.port=8545 --authrpc.port=8551 --torrent.port=42069 --private.api.addr= --http --ws --http.api=eth,debug,net,trace,web3,erigon

# sepolia
./build/bin/erigon --datadir="<your_sepolia_data_path>" --chain=sepolia --port=30304 --http.port=8546 --authrpc.port=8552 --torrent.port=42068 --private.api.addr= --http --ws --http.api=eth,debug,net,trace,web3,erigon

Quote your path if it has spaces.

Dev Chain

🔬 Detailed explanation is DEV_CHAIN.

Key features

🔬 See more detailed overview of functionality and current limitations. It is being updated on recurring basis.

More Efficient State Storage

Flat KV storage. Erigon uses a key-value database and storing accounts and storage in a simple way.

🔬 See our detailed DB walkthrough here.

Preprocessing. For some operations, Erigon uses temporary files to preprocess data before inserting it into the main DB. That reduces write amplification and DB inserts are orders of magnitude quicker.

🔬 See our detailed ETL explanation here.

Plain state.

Single accounts/state trie. Erigon uses a single Merkle trie for both accounts and the storage.

Faster Initial Sync

Erigon uses a rearchitected full sync algorithm from Go-Ethereum that is split into "stages".

🔬 See more detailed explanation in the Staged Sync Readme

It uses the same network primitives and is compatible with regular go-ethereum nodes that are using full sync, you do not need any special sync capabilities for Erigon to sync.

When reimagining the full sync, with focus on batching data together and minimize DB overwrites. That makes it possible to sync Ethereum mainnet in under 2 days if you have a fast enough network connection and an SSD drive.

Examples of stages are:

  • Downloading headers;

  • Downloading block bodies;

  • Recovering senders' addresses;

  • Executing blocks;

  • Validating root hashes and building intermediate hashes for the state Merkle trie;

  • [...]

JSON-RPC daemon

Most of Erigon's components (txpool, rpcdaemon, snapshots downloader, sentry, ...) can work inside Erigon and as independent process.

To enable built-in RPC server: --http and --ws (sharing same port with http)

Run RPCDaemon as separated process: this daemon can use local DB (with running Erigon or on snapshot of a database) or remote DB (run on another server). 🔬 See RPC-Daemon docs

For remote DB

This works regardless of whether RPC daemon is on the same computer with Erigon, or on a different one. They use TPC socket connection to pass data between them. To use this mode, run Erigon in one terminal window

make erigon
./build/bin/erigon --private.api.addr=localhost:9090 --http=false
make rpcdaemon
./build/bin/rpcdaemon --private.api.addr=localhost:9090 --http.api=eth,erigon,web3,net,debug,trace,txpool

gRPC ports

9090 erigon, 9091 sentry, 9092 consensus engine, 9093 torrent downloader, 9094 transactions pool

Supported JSON-RPC calls (eth, debug , net, web3):

For a details on the implementation status of each command, see this table.

Run all components by docker-compose

Docker allows for building and running Erigon via containers. This alleviates the need for installing build dependencies onto the host OS.

Optional: Setup dedicated user

User UID/GID need to be synchronized between the host OS and container so files are written with correct permission.

You may wish to setup a dedicated user/group on the host OS, in which case the following make targets are available.

# create "erigon" user
make user_linux
# or
make user_macos

Environment Variables

There is a .env.example file in the root of the repo.

  • DOCKER_UID - The UID of the docker user
  • DOCKER_GID - The GID of the docker user
  • XDG_DATA_HOME - The data directory which will be mounted to the docker containers

If not specified, the UID/GID will use the current user.

A good choice for XDG_DATA_HOME is to use the ~erigon/.ethereum directory created by helper targets make user_linux or make user_macos.

Check: Permissions

In all cases, XDG_DATA_HOME (specified or default) must be writeable by the user UID/GID in docker, which will be determined by the DOCKER_UID and DOCKER_GID at build time.

If a build or service startup is failing due to permissions, check that all the directories, UID, and GID controlled by these environment variables are correct.


Next command starts: Erigon on port 30303, rpcdaemon on port 8545, prometheus on port 9090, and grafana on port 3000.

# Will mount ~/.local/share/erigon to /home/erigon/.local/share/erigon inside container
make docker-compose

# or
# if you want to use a custom data directory
# or, if you want to use different uid/gid for a dedicated user
# To solve this, pass in the uid/gid parameters into the container.
# DOCKER_UID: the user id
# DOCKER_GID: the group id
# XDG_DATA_HOME: the data directory (default: ~/.local/share)
# Note: /preferred/data/folder must be read/writeable on host OS by user with UID/GID given
#       if you followed above instructions
# Note: uid/gid syntax below will automatically use uid/gid of running user so this syntax
#       is intended to be run via the dedicated user setup earlier
DOCKER_UID=$(id -u) DOCKER_GID=$(id -g) XDG_DATA_HOME=/preferred/data/folder DOCKER_BUILDKIT=1 COMPOSE_DOCKER_CLI_BUILD=1 make docker-compose

# if you want to run the docker, but you are not logged in as the $ERIGON_USER
# then you'll need to adjust the syntax above to grab the correct uid/gid
# To run the command via another user, use

Makefile creates the initial directories for erigon, prometheus and grafana. The PID namespace is shared between erigon and rpcdaemon which is required to open Erigon's DB from another process (RPCDaemon local-mode). See:

If your docker installation requires the docker daemon to run as root (which is by default), you will need to prefix the command above with sudo. However, it is sometimes recommended running docker (and therefore its containers) as a non-root user for security reasons. For more information about how to do this, refer to this article.

Windows support for docker-compose is not ready yet. Please help us with .ps1 port.

Grafana dashboard

docker compose up prometheus grafana, detailed docs.

old data

Disabled by default. To enable see ./build/bin/erigon --help for flags --prune


The ./docs directory includes a lot of useful but outdated documentation. For code located in the ./cmd directory, their respective documentation can be found in ./cmd/*/ A more recent collation of developments and happenings in Erigon can be found in the Erigon Blog.


How much RAM do I need

  • Baseline (ext4 SSD): 16Gb RAM sync takes 6 days, 32Gb - 5 days, 64Gb - 4 days
  • +1 day on "zfs compression=off". +2 days on "zfs compression=on" (2x compression ratio). +3 days on btrfs.
  • -1 day on NVMe

Detailed explanation: ./docs/programmers_guide/

Default Ports and Firewalls

erigon ports

Component Port Protocol Purpose Should Expose
engine 9090 TCP gRPC Server Private
engine 42069 TCP & UDP Snap sync (Bittorrent) Public
engine 6060 TCP Metrics or Pprof Private
engine 8551 TCP Engine API (JWT auth) Private
sentry 30303 TCP & UDP eth/68 peering Public
sentry 30304 TCP & UDP eth/67 peering Public
sentry 9091 TCP incoming gRPC Connections Private
rpcdaemon 8545 TCP HTTP & WebSockets & GraphQL Private

Typically, 30303 and 30304 are exposed to the internet to allow incoming peering connections. 9090 is exposed only internally for rpcdaemon or other connections, (e.g. rpcdaemon -> erigon). Port 8551 (JWT authenticated) is exposed only internally for Engine API JSON-RPC queries from the Consensus Layer node.

caplin ports

Component Port Protocol Purpose Should Expose
sentinel 4000 UDP Peering Public
sentinel 4000 UDP Peering Public
sentinel 4001 TCP Peering Public
sentinel 7777 TCP gRPC Connections Private

If you are using --internalcl aka caplin as your consensus client, then also look at the chart above

shared ports

Component Port Protocol Purpose Should Expose
all 6060 TCP pprof Private
all 6060 TCP metrics Private

Optional flags can be enabled that enable pprof or metrics (or both) - however, they both run on 6060 by default, so

you'll have to change one if you want to run both at the same time. use --help with the binary for more info.

other ports

Reserved for future use: gRPC ports: 9092 consensus engine, 9093 snapshot downloader, 9094 TxPool

Hetzner expecting strict firewall rules             "This" Network             RFC 1122, Section            Private-Use Networks       RFC 1918         Carrier-Grade NAT (CGN)    RFC 6598, Section 7         Private-Use Networks       RFC 1918        Link Local                 RFC 3927         Private-Use Networks       RFC 1918          IETF Protocol Assignments  RFC 5736          TEST-NET-1                 RFC 5737        6to4 Relay Anycast         RFC 3068        Private-Use Networks       RFC 1918         Network Interconnect
Device Benchmark Testing   RFC 2544       TEST-NET-2                 RFC 5737        TEST-NET-3                 RFC 5737           Multicast                  RFC 3171           Reserved for Future Use    RFC 1112, Section 4    Limited Broadcast          RFC 919, Section 7
RFC 922, Section 7

Same in IpTables syntax

How to get diagnostic for bug report?

  • Get stack trace: kill -SIGUSR1 <pid>, get trace and stop: kill -6 <pid>
  • Get CPU profiling: add --pprof flag run go tool pprof -png\?seconds\=20 > cpu.png
  • Get RAM profiling: add --pprof flag run go tool pprof -inuse_space -png > mem.png

How to run local devnet?

🔬 Detailed explanation is here.

Docker permissions error

Docker uses user erigon with UID/GID 1000 (for security reasons). You can see this user being created in the Dockerfile. Can fix by giving a host's user ownership of the folder, where the host's user UID/GID is the same as the docker's user UID/GID (1000). More details in post

Run RaspberyPI

How to change db pagesize


Getting in touch

Erigon Discord Server

The main discussions are happening on our Discord server. To get an invite, send an email to bloxster [at] with your name, occupation, a brief explanation of why you want to join the Discord, and how you heard about Erigon.

Reporting security issues/concerns

Send an email to security [at]

Known issues

htop shows incorrect memory usage

Erigon's internal DB (MDBX) using MemoryMap - when OS does manage all read, write, cache operations instead of Application (linux , windows)

htop on column res shows memory of "App + OS used to hold page cache for given App", but it's not informative, because if htop says that app using 90% of memory you still can run 3 more instances of app on the same machine - because most of that 90% is "OS pages cache". OS automatically frees this cache any time it needs memory. Smaller "page cache size" may not impact performance of Erigon at all.

Next tools show correct memory usage of Erigon:

  • vmmap -summary PID | grep -i "Physical footprint". Without grep you can see details

    • section MALLOC ZONE column Resident Size shows App memory usage, section REGION TYPE column Resident Size shows OS pages cache size.
  • Prometheus dashboard shows memory of Go app without OS pages cache (make prometheus, open in browser localhost:3000, credentials admin/admin)

  • cat /proc/<PID>/smaps

    Erigon uses ~4Gb of RAM during genesis sync and ~1Gb during normal work. OS pages cache can utilize unlimited amount of memory.

    Warning: Multiple instances of Erigon on same machine will touch Disk concurrently, it impacts performance - one of main Erigon optimisations: "reduce Disk random access". "Blocks Execution stage" still does many random reads - this is reason why it's slowest stage. We do not recommend running multiple genesis syncs on same Disk. If genesis sync passed, then it's fine to run multiple Erigon instances on same Disk.

Blocks Execution is slow on cloud-network-drives

Please read In short: network-disks are bad for blocks execution - because blocks execution reading data from db non-parallel non-batched way.

Filesystem's background features are expensive

For example: btrfs's autodefrag option - may increase write IO 100x times

Gnome Tracker can kill Erigon

Gnome Tracker - detecting miners and kill them.

the --mount option requires BuildKit error

For anyone else that was getting the BuildKit error when trying to start Erigon the old way you can use the below...

XDG_DATA_HOME=/preferred/data/folder DOCKER_BUILDKIT=1 COMPOSE_DOCKER_CLI_BUILD=1 make docker-compose