This project is a node.js client library for the Weed-FS REST interface.
Weed-FS is a simple and highly scalable distributed file system. It focuses on two objectives:
- storing billions of files!
- and serving them fast!
Weed-FS chose to implement only a key~file mapping instead of supporting full POSIX file system semantics. This can be called "NoFS". (Similar to "NoSQL")
Instead of managing all file metadata in a central master, Weed-FS manages file volumes in the central master, and allows volume servers to manage files and the metadata. This relieves concurrency pressure from the central master and spreads file metadata into memory on the volume servers allowing faster file access with just one disk read operation!
Weed-FS models after Facebook's Haystack design paper and costs only 40 bytes disk storage for each file's metadata. It is so simple with O(1) disk read that anyone is more than welcome to challenge the performance with actual use cases.
var weedClient = require("weed-fs");
var weedfs = new weedClient({
server: "localhost",
port: "9333"
});
weedfs.write("./file.png", function(err, fileInfo) {
if (err) {
console.error(err);
} else {
console.log(fileInfo);
}
});Use the write() function to store files. The callback recieves the parsed JSON response.
Anything passed to the {opts} is made into a query string and
is used with the /dir/assign HTTP request. You can use this to define the replication strategy.
client.write("./file.png", {replication: 000}, function(err, fileInfo) {
if (fileInfo.error) {
throw fileInfo.error;
}
console.log(fileinfo);
});You can also write multiple files:
client.write(["./fileA.jpg", "./fileB.jpg"], function(err, fileInfo) {
// This callback will be called for both fileA and fileB.
// The fid's will be the same, to access each variaton just
// add _ARRAYINDEX to the end of the fid. In this case fileB
// would be: fid + "_1"
var fidA = fileInfo;
var fidB = fileInfo + "_1";
console.log(fileInfo);
}For both methods of writing, a stream can be passed as a file.
The read function supports streaming. To use simply do:
client.read(fileId, fs.createWriteStream("read.png"));If you prefer not to use streams just use:
client.read(fileId, function(err, response, body) {
if (err) {
throw err;
}
// Here's your data:
var filedata = body;
});This function can be used to find the location(s) of a file amongst the cluster.
client.find(fileId, function(public, servers) {
console.log(public[0]);
// servers contains the non-public URLs. Use this for editing and removing.
});This function will delete a file from the store. If server is specified, the
file will only be removed from that location. Otherwise it will be deleted from all locations.
client.remove(fileId, function(err, resp, body) {
if (err) {
throw err;
}
console.log("removed file.");
});This function will query the master status for status information. The callback contains an object containing the information.
client.systemStatus(function(status) {
console.log(status);
});This function will query an individual volume server for server-specific information.
client.status("localhost", 8080, function(status) {
console.log(status);
});This function will force the master server to preform garbage collection on volume servers.
If your system has many deletions, the deleted file's disk space will not be synchronously re-claimed. There is a background job to check volume disk usage. If empty space is more than the threshold, default to 0.3, the vacuum job will make the volume readonly, create a new volume with only existing files, and switch on the new volume. If you are impatient or doing some testing, vacuum the unused spaces this way.
client.vacuum({garbageThreshold: 0.4}, function(status) {
console.log(status);
});Copyright (C) 2013-2022 Aaron Blakely