mdbx-remote is a tool to access your MDBX databases over network safely without any concern of corrupting the databases or getting dirty data. mdbx-remote also provides the same but async programming interface as the original libmdbx-rs bindings, which passes all tests of origin libmdbx-rs.

MDBX is a fast and easy-to-use KV database offering ACID properties. However, one of the limitations of MDBX is that it tries hard to prevent users from accessing the database via network to avoid potential corruption. For example, opening a database in the common collaborative mode will immediately get MDBX_EREMOTE. Even if you open the database via MDBX_EXCLUSIVE, it is still highly likely to corrupt the database or get inconsistent data if there are remote readers, which MDBX can not always detect. This is super inconvenient in many cases, like just changing one or two entries.

mdbx-remote solves this by spinning up an RPC service to expose all raw MDBX primitives.

Simply do

cargo build --release

Note libclang-dev is required to generate the libmdbx bindings.

Alternatively, there is a docker image provided to help setup.

mdbx-remote is shipped with an mdbx utlity. For example, spin up an RPC endpoint:

# By default, this will listen at 0.0.0.0:1899
RUST_LOG=mdbx=info cargo run --release server
# Or 
docker run --rm -it --env RUST_LOG=info lazymio/mdbx-remote server

And stat any database with a url:

# Stat a remote reth database
cargo run --release stat mdbx://127.0.0.1/remote/path/to/db?ro=1
# Or
docker run --rm -it --env RUST_LOG=info --network host \
    lazymio/mdbx-remote stat mdbx://127.0.0.1/remote/path/to/db?ro=1

Database:
	Page size: 4096
	Entries: 19265135020
	Leaf Pages: 365773779
	Branch Pages: 23212389
	Depth: 102
	Overflow Pages: 5270586
Table Main:
	Page size: 4096
	Entries: 29
	Leaf Pages: 1
	Branch Pages: 0
	Depth: 1
	Overflow Pages: 0
Table AccountChangeSets:
	Page size: 4096
	Entries: 3298505382
	Leaf Pages: 42669897
	Branch Pages: 13964000
	Depth: 4
	Overflow Pages: 0
...

This exactly mimics the output of mdbx_stat utility from libmdbx. Note the mdbx:// URL supports a subset of MDBX features by parameters including:

• ro: Read-only
• rw: Read-write
• accede: Use options from existing environment
• exclusive: Open in the exclusive mode
• no_sub_dir: No directory for the mdbx database
• max_readers: Max readers
• max_dbs: Max databases
• deadline: The seconds to wait before giving up an RPC request

It is also worth mentioning that, if the host part is empty, like mdbx:///path/to/local/db, it refers to a local MDBX database located at /path/to/local/db.

One of the motivation to build mdbx-remote is to access reth database. mdbx reth can get a value from one reth table.

cargo run --release reth \
    -u mdbx://127.0.0.1/path/to/reth/db \
    -t PlainAccountState \
    0x95222290DD7278Aa3Ddd389Cc1E1d165CC4BAfe5
# Or
docker run --rm -it --network host lazymio/mdbx-remote reth \
    -u mdbx://127.0.0.1/path/to/reth/db \
    -t PlainAccountState \
    0x95222290DD7278Aa3Ddd389Cc1E1d165CC4BAfe5

The account is:
{
  "nonce": 1569575,
  "balance": "0x7db63f1578c045b9",
  "bytecode_hash": null
}

mdbx-remote reaches almost the full parity with libmdbx-rs interfaces, except forcing everything to async because of tarpc. Generally, the table below shows the types for different scenarios.

Add this to your Cargo.toml:

libmdbx-remote = "0.3"

You certainly don't want to make 1 RPC call per data fetching, which wastes your bandwith and is super slow.

mdbx-remote provides buffered cursor operations for high performance data accessing.

Check the methods with _buffered suffix for details.

• Only expose the RPC endpoint to trusted network because it doesn't have any protection or authentication. Tunnel the traffic via something like SSH if you really would like to expose it to public network.
• The same database can't be opened twice as inherited from MDBX.

• Batch cursor (taking a self) get/put to improve throughput
• Implmenet more MDBX utilities, especially mdbx_dump.
• Reach full parity with libmdbx-rs.
• Reach more parity with reth db operations.
• Benchmark the performance.
• Run Reth with remote storage.

The bindings is forked from reth v1.1.0.