dynalite

dynalite

An implementation of Amazon's DynamoDB, focussed on correctness and performance, and built on LevelDB (well, @rvagg's awesome LevelUP to be precise).

This project aims to match the live DynamoDB instances as closely as possible (and is tested against them in various regions), including all limits and error messages.

Why not Amazon's DynamoDB Local?

Because it's too buggy! And it differs too much from the live instances in a number of key areas (see below)

Example

$ dynalite --help

Usage: dynalite [--port <port>] [--path <path>] [options]

A DynamoDB http server, optionally backed by LevelDB

Options:
--help                Display this help message and exit
--port <port>         The port to listen on (default: 4567)
--path <path>         The path to use for the LevelDB store (in-memory by default)
--ssl                 Enable SSL for the web server (default: false)
--createTableMs <ms>  Amount of time tables stay in CREATING state (default: 500)
--deleteTableMs <ms>  Amount of time tables stay in DELETING state (default: 500)
--updateTableMs <ms>  Amount of time tables stay in UPDATING state (default: 500)

Report bugs at github.com/mhart/dynalite/issues

Or programmatically:

// Returns a standard Node.js HTTP server
var dynalite = require('dynalite'),
    dynaliteServer = dynalite({path: './mydb', createTableMs: 50})

// Listen on port 4567
dynaliteServer.listen(4567, function(err) {
  if (err) throw err
  console.log('Dynalite started on port 4567')
})

Installation

With npm do:

$ npm install -g dynalite

TODO

• Add ProvisionedThroughput checking
• Add config settings to turn on/off strict checking
• Use efficient range scans for Query calls
• Explore edge cases with Query and ScanIndexForward: false (combine with above)
• Implement ReturnItemCollectionMetrics on all remaining endpoints
• Implement size info for tables and indexes
• Is the ListTables limit of names returned 100 if no Limit supplied?

Problems with Amazon's DynamoDB Local

Part of the reason I wrote dynalite was due to the existing mock libraries not exhibiting the same behaviour as the live instances. Amazon released their DynamoDB Local Java tool recently, but the current version (2013-09-12) still has quite a number of issues that have prevented us (at Adslot) from testing our production code, especially in a manner that simulates actual behaviour on the live instances.

Some of these are documented (eg, no ConsumedCapacity returned), but most aren't - the items below are a rough list of the issues we've found (and do not exist in dynalite), vaguely in order of importance:

• Returns 400 when UpdateItem uses the default PUT Action without explicitly specifying it (this actually prevents certain client libraries from being used at all)
• Does not return correct number of UnprocessedKeys in BatchGet (returns one less!)
• Returns 400 when trying to put valid numbers with less than 38 significant digits, eg 1e40
• Returns 200 for duplicated keys in BatchGetItem
• Returns 200 when hash key is too big in GetItem/BatchGetItem
• Returns 200 when range key is too big in GetItem/BatchGetItem
• Returns 200 for PutItem/GetItem/UpdateItem/BatchGetItem/Scan/etc with empty strings (eg, {a: {S: ''}})
• Returns 413 when request is over 1MB (eg, in a BatchWrite with 25 items of 64k), but live instances allow 8MB
• Returns ResourceNotFoundException in ListTables if ExclusiveStartName no longer exists
• Does not return ConsistentRead property in UnprocessedKeys in BatchGet even if requested
• Returns 200 for empty RequestItems in BatchGetItem/BatchWriteItem
• Returns 200 when trying to delete NS from SS or NS from N or add NS to SS or N to NS
• Allows UpdateTable when read and write capacity are same as current (should be an error)
• Tables are created in ACTIVE state, not CREATING state
• Tables are removed without going into a DELETING state
• Tables are updated without going into a UPDATING state
• PutItem returns Attributes by default, even though none are requested
• Does not add ProvisionedThroughput.LastIncreaseDateTime in UpdateTable
• Does not update ProvisionedThroughput.NumberOfDecreasesToday in UpdateTable
• Different serialization and validation error messages from live instances (makes it hard to debug)
• Uses uppercase Message for error messages (should only use uppercase for SerializationException)
• Often returns 500 instead of 400 (or similarly appropriate status)
• Doesn't return ConsumedCapacity (documented - but makes it very hard to calculate expected usage)
• Does not calculate the Scan size limits correctly so can return too many items
• Does not return LastEvaluatedKey on a Query when at end of table
• Does not return LastEvaluatedKey on a Scan when Limit equals number of matching items
• Does not return X-Amzn-RequestId header
• Does not return X-Amz-Crc32 header
• Does not return application/json if application/json is requested
• Fails to put numbers 1e-130 and -1e-130 (succeeds on live instances)
• Returns an error when calling Query on a hash table (succeeds on live instances)
• Returns 500 if random attributes are supplied (just ignored on live instances)
• Does not convert doubles to booleans (returns 500 instead)