creditkarma / memcached Goto Github PK

See 3rd-Eden#342 for the original, the same issue exists in this fork. I've repasted the contents below.

If you get in a state where memcached attempts to reconnect, it will call ping(this.tokens[1]), which attempts to spawn a binary to call ping -c 3. The ping function in Javascript does not handle the case where the ping binary is not present - specifically it's missing an .on(error) handler that could catch the situation where ping is not found, has the wrong permissions, or the running process does not have permission to spawn children.

function ping (host, callback) {
  var isWin = process.platform.indexOf('win') === 0; // win32 or win64
  var arg = isWin ? '-n' : '-c';
  var pong = spawn('ping', [arg, '3', host]); // only ping 3 times

  pong.stdout.on('data', function stdoutdata (data) {
    callback(false, data.toString().split('\n')[0].substr(14));
    pong.kill();
  });

  pong.stderr.on('data', function stderrdata (data) {
    callback(new Error(data.toString().split('\n')[0].substr(14)), false);
    pong.kill();
  });
}

The lack of a .on('error') handler means that any failure here is uncatchable and crashes the process.

More broadly I have to wonder about the effectiveness of calling ping if a connection fails - it's not clear that one TCP packet is going to have any more luck than another if the connection is in trouble.

I'm facing the same issue as described here: 3rd-Eden#339

I'm still using 3rd-Eden package, but worth the shot asking here if you know how this issue could be solved.

My current config is and I don't run in cluster mode:

const memcached = new Memcached(MEMCACHED_ADDRESS, {
  maxExpiration: MAX_EXPIRATION_TIME,
  maxValue: 2097152, // Max length of cache value in bytes
  timeout: 3000, // after x ms the server should send a timeout if we can't connect
  poolSize: 50, // maximal parallel connections
  retries: 2, // Connection pool retries to pull connection from pool
  minTimeout: 500, // Connection pool retry min delay before retrying
  maxTimeout: 1000, // Connection pool retry max delay before retrying
  retry: 1000, // When a server has an error, wait this amount of time before retrying
})

I'm also wondering if this was a 1-to-1 rewrite or did you guys fix any bugs and are going to maintain this lib?

It is my interest to update to a lib that at least is minimally maintained