Fetching contributors…
Cannot retrieve contributors at this time
160 lines (117 sloc) 5.39 KB
Membase Command Line Tool Set Specification
High Level Description
There will be a command line tool set that exists that allows users to have
the same access to functions such as:
* Starting and stopping of the server
* The dynamic cluster membership changes of the
Membase Server
* Query/fetch a given key
* Joining a host:port
* Listing nodes
* Replication information
* Set a given key/value
* Not only support the Membase Server, but
could also support the Membase Memcached Server since both
of these products have a well-defined streaming REST
(Representational State Transfer) interface.
* Be able to point to any server node in
the cluster to be able to manage the entire Membase cluster.
* Be extensible. To add new functions or commands to the tool,
there needs to be a way to easy add new functions with the least
amount of changes to the overall tool as well as existing tools
Mid-level Discription
These tools will have some primary components:
* Use the chain of responsibility design pattern. A good example of
this can be found at:
Although this example is in C++, the basic concept will be possible
to be implemented in Python.
* These tools will be written in Python, using a framework which works best
for extensibility
* Top level source file:
- parses user input to obtain
* operation name
* command line flags specific to that command of the desired operation.
- has dictionary of functions keyed by the name which will be a
"dispatch hash"
* Source file with the implementation of functions for sending REST
requests to the specified server and processing the JSON response.
- PycURL library will provide the HTTP client functionality
- Parsing of JSON into objects will be processed by the json module.
The ouput of this processing can be used by the command line
to display the JSON results of the REST command in a printable format
For instance:
* membase.py - contains parsing of command, dictionary of functions keyed by
command string to dispatch appropriate method.
* membase_cli_list_servers.py - contains the function that implements the
servers listing functionality
* membase_cli_$operation - other operations, as they are implemented
* membase_cli_rest_client.py - contains the functions for :
- converting json replies from the REST interface to something that
can be printed out by the program for the given command
- HTTP client interface for sending REST commands and recieving
the JSON replies
The following are operations that each function source file will implement:
command as run by user REST command
membase listservers GET /pools/default
membase join POST /node/controller/doJoinCluster
-c cluster (IP:port, host:port)
-u user
-p password
membase eject POST /node/controller/ejectNode
-c cluster (IP:port, host:port)
-u user
-p password
membase listpools GET /pools/default
-s servers [server,]
membase listbuckets GET /pools/default/buckets/default
-p pool [pool,]
membase flush GET /pools/default/bucket/default/controller/doFlush
membase stats GET /pools/default/stats?stat=$stat&period=$period
-p pools [pool,]
-s servers [server,]
-b buckets
membase bucket POST /pools/$pool/buckets (pool != default)
-c create
-n bucketname name=$bucketname
-g get
-p pool
-d delete
-f flush [bucket,]
-s cachesize cacheSize=$cachesize
membase log
-g get
-e entry text
membase version
Low-level Discription
Chain of responsibility/dispatcher
One crude idea is a big switch() statement to call the appropriate
command or function. But the article above brings to light a more
flexible and extensible model: using a Python dictionary of functions,
keyed by the operation for whatever functions exist and to simple call
the appropriate function/method based on the command (hash key, value
being the function/method), without having to use a switch or
interpolation of the command string into a numeric value tested by a
switch statement.
Each command implementation will exist in its own source file, and
whatever happens to be imported into the top-level program, that
command should simply run as it is keyed from the function dictionary.
Networking Layer using REST
The source file membase_cli_rest_client.py will contain the
functions for sending REST requests to the server using PyCURL.
Imported by each source file implementing a particular command, the
programs that implment each REST action will have access to these
JSON parsing of REST replies
membase_cli_rest_client.py will also contain the functions/methods
for parsing the REST JSON replies into data objects that can be
displayed to the user.
Command line parsing
The top level source file, membase.py, will contain the functionality for
parsing user input to the main command line program.