lapidus

Stream your PostgreSQL, MySQL or MongoDB databases anywhere, fast.

Showing:

Popularity

Downloads/wk

2

GitHub Stars

154

Maintenance

Last Commit

8mos ago

Contributors

4

Package

Dependencies

12

Size (min+gzip)

16.4KB

License

MIT

Type Definitions

Tree-Shakeable

No?

Categories

Readme

Lapidus

Lapidus

Build Status Coverage Status Join the chat at https://gitter.im/JarvusInnovations/lapidus

Getting Started

Currently MySQL and PostgreSQL databases are fully supported. MongoDB supports inserts and deletes, however, updates return a copy of the operation (for example, a $rename operation will return a $set for the new field and an $unset for the old field) instead of the object as it exists in the database. Redis support is on the way. Lapidus can currently be used as a daemon or Node.js module. Support for piping line-delimited JSON to other processes is a high priority.

To install

npm install -g lapidus

PostgreSQL

You'll need PostgreSQL 9.4 or higher with logical replication configured and the JSONCDC plugin installed and loaded. Any PostgreSQL fork that ships with pg_recvlogical should be compatible.

To install the JSONCDC logical decoding plugin using pgxn:

sudo easy_install pgxnclient
pgxn install jsoncdc --testing

NOTE: JSONCDC also provides .debs inside their releases repo, if you wish to not install through pgxn.

To enable logical decoding and the JSONCDC plugin add the following lines to your postgresql.conf:

wal_level = logical
max_wal_senders = 10
max_replication_slots = 10

shared_preload_libraries = 'jsoncdc'

Create a user with replication privileges and add them to your pg_hba.conf file.

Afterwards, restart PostgreSQL and verify that it starts correctly:

service postgresql restart
service postgresql status

WARNING: PostgreSQL will hold onto the WAL logs until all logical replication slots have consumed their data. This means that if you try out Lapidus and fail to delete your slot that you'll likely run out of disk space on your system.

For information on managing replication slots: consult the documentation.

MySQL

You'll need MySQL 5.1.15 or higher with binary logging configured. Forks of MySQL should be compatible but have not been tested.

Add the following lines to your my.cnf:

server-id        = 1
log_bin          = /var/log/mysql/mysql-bin.log
max_binlog_size  = 100M  # WARNING: make sure to set this to a sane value or you may fill your disk
expire_logs_days = 10    # Optional
binlog_format    = row

Create a user with replication permissions and select permissions:

GRANT REPLICATION SLAVE, REPLICATION CLIENT, SELECT ON *.* TO 'lapidus'@'localhost' IDENTIFIED BY 'secure-password';

Restart MySQL and verify that it starts correctly:

service mysql restart
service mysql status

MongoDB

We test against MongoDB 3.x, however, older versions should work. You'll need to setup MongoDB as a replica set. If you're not truly using replication during development you will need to connect and run:

// DO NOT DO THIS IN PRODUCTION
rs.initiate()
db.getMongo().setSlaveOk();

For more information on setting up replication in MongoDB check out the docs.

Configuration

Lapidus will search for lapidus.json in the current working directory. You can specify a different configuration by passing it to the constructor or using the -c flag on the terminal. For a list of command line options run lapidus --help.

Here is a sample configuration file that will connect to two PostgreSQL backends, two MySQL backends, one MongoDB backend and publish all events to NATS using the NATS plugin:

{
  "backends": [
    {
      "type": "mysql",
      "hostname": "127.0.0.1",
      "username": "jacob",
      "database": "jacob",
      "password": "2PQM9aiKMJX5chv76gYdFJNi",
      "serverId": 1,
      "excludeTables": [
        "sessions"
      ]
    },

    {
      "type": "mysql",
      "hostname": "127.0.0.1",
      "username": "sawyer",
      "database": "sawyer",
      "password": "2PQM9aiKMJX5chv76gYdFJNi",
      "serverId": 2
    },

    {
      "type": "postgresql",
      "host": "127.0.0.1",
      "user": "lapidus",
      "database": "lapidus",
      "password": "2PQM9aiKMJX5chv76gYdFJNi",
      "slot": "lapidus_slot"
    },

    {
      "type": "postgresql",
      "host": "127.0.0.1",
      "user": "hurley",
      "database": "hurley",
      "password": "2PQM9aiKMJX5chv76gYdFJNi",
      "slot": "hurley_slot"
    },

    {
      "type": "mongo",
      "hostname": "127.0.0.1",
      "username": "lapidus",
      "database": "lapidus",
      "password": "2PQM9aiKMJX5chv76gYdFJNi",
      "replicaSet": "rs0"
    }
  ],
  "plugins": {
    "nats": {
      "server": "nats://localhost:4222"
    }
  }
}

Plugins

NATS

Lapidus ships with a lightweight NATS plugin. NATS is an open-source, high-performance, lightweight cloud native messaging system.

###Configuration

Publish to NATS for all backends (one connection per backend):

{
  "backends": [...]
  ],
  "plugins": {
    "nats": {
      "server": "nats://localhost:4222"
    }
  }
}

Publish to NATS for a specific backend:

{
  "backends": [
    {
      "type": "postgresql",
      "host": "the.hatch",
      "user": "desmond",
      "database": "darma",
      "password": "notpennysboat123",
      "slot": "walts_raft",

      "plugins": {
        "nats": {
          "server": "nats://localhost:4222"
        }
      }
    }
  ]
}

Events

Insert, Update and Delete events will be published using the subject schema.table.pk. Here are examples events:

Insert:

// TODO: sample insert

Update:

// TODO: sample update

Delete:

// TODO: sample delete

Production usage

Lapidus worked well for 5k concurrent clients on $5-10 Digital Ocean droplets using the MySQL and PostgreSQL backends and per-tenant socket.io servers using the NATS plugin. Typical latency between MySQL -> Lapidus -> NATS within a datacenter was 1ms - 3ms. Since our messages were fire-and-forget, we never solved for the gotchas below and our end-user lag and CPU usage were neglible . Since NATS will disconnect slow consumers there was good tenant isolation without much work.

Gotchas

To prevent a deluge of events during bulk loading of data or restoration of backups make sure your procedures/scripts stop Lapidus and/or delete/reset any persistent subscriptions or message queues external to Lapdidus as needed.

Resource requirements

CPU

CPU usage is light, as a rule of thumb, measure your peak MySQL CPU usage (after enabling binary logging) and multiply that by 0.075. That's how much CPU Lapidus is likely to use at peak.

Memory

Generally speaking, each worker requires 10-15 MB of ram.

Your peak memory usage is dictated by V8's garbage collection. When running the TPC-C benchmark against the MySQL worker using 8 cores memory sat around 70MB and peaked at 120MB before garbage collection knocked it back down to 70MB.

I tested for memory leaks by running 5 million transactions using the TPC-C benchmark and things look pretty solid, if you notice any issues please report them.

License

Lapidus is MIT licensed. The artwork in the header is Copyright Matt Greenholt.

Contributors

Matt Greenholt has kindly allowed the use of his artwork. Check out his blog and flickr.

Rate & Review

Great Documentation0
Easy to Use0
Performant0
Highly Customizable0
Bleeding Edge0
Responsive Maintainers0
Poor Documentation0
Hard to Use0
Slow0
Buggy0
Abandoned0
Unwelcoming Community0
100
No reviews found
Be the first to rate

Alternatives

No alternatives found

Tutorials

No tutorials found
Add a tutorial