db-oracle is not flushing data

Question

I'm working on a simple service using Node.js. It receives uploaded files, stores them on the disk and records some metadata on an Oracle table. I'm using the db-oracle package together with connection pooling, following this article: http://nodejsdb.org/2011/05/connection-pooling-node-db-with-generic-pool/

However, I've noticed that the data I insert is only sent to the Oracle database after the connection pool closes the idle connection, by calling its disconnect() method.

Is there a way to flush the data before sending the 'OK' signal to my client? The way it is working now, a crash on my webservice or on Oracle itself can cause loss of data, and the client of my service would not know about it. I actually tested this by killing my app process after some uploads, and the data was indeed lost.

Here's a simplified version of the code:

var express = require('express');
var app = module.exports = express.createServer();

app.post('/upload', handleUpload);

app.listen(4001, function(){
  console.log("Express server listening on port %d in %s mode", app.address().port, app.settings.env);
});

function handleUpload(req, res) {
  res.contentType('application/xml');

  var buf = '';
  req.on('data', function(chunk) { buf += chunk; });
  req.on('end', function() {
    saveUpload(req, res, buf);
  });
}

function saveUpload(req, res, buf) {
  if (buf.length == 0)
    return sendError(res, 'No data supplied', 422);

  var payload = new Buffer(buf, 'base64');

  files.save(payload, function(err, savedFile) {
    if (err)
      return sendError(res, 'Error while saving', 500);

    var obj = { ip: req.connection.remoteAddress, location: savedFile.path,
                created_at: new Date(), updated_at: new Date() };

    var fields = ['IP', 'LOCATION', 'CREATED_AT', 'UPDATED_AT'];
    var values = fields.map(function(v) { return obj[v.toLowerCase()] });

    pool.acquire(function(err, conn) {
      if (err)
        return sendError(res, err, 500);

      var q = conn.query().insert('FILES', fields, values);

      q.execute(function(err, result) {
        pool.release(conn);

        if (err)
          return sendError(res, err, 500);

        if (result.affected < 1)
          return sendError(res, 'Error saving the record', 500);

        // The next statement sends the final result to the client.
        // However, the new record was not yet flushed to the database.
        res.end('<ok />');
      });
    });
  });
}

function sendError(res, err, code) {
  console.log(err);
  res.send('<error>' + err + '</error>', code || 500);
}

As a workaround, I've tried to implement a fake connection pool and release all acquired connections, but now my app is dying with the message: pure virtual method calledAbort trap: 6

Here's the fake connection pooling:

var fakePool = {
  acquire: function(callback) {
    new oracle.Database(config.database).connect(function(err, server) {
      callback(err, this);
    });
  },
  release: function(conn) {
    conn.disconnect();
  }
};

Just to be clear, I don't care about the fake connection pooler, it was just a dirty workaround. I want to be able to flush the data to Oracle before sending the 'OK' to my client.

Btw I also opened a ticket on their Github: https://github.com/mariano/node-db-oracle/issues/38

Answer 1

You are obviously missing a transaction commit.

node-db does not need to expose a commit API because in most RDBMS (including Oracle), COMMIT is a valid query. Since the package allows the execution of arbitrary queries, commit/rollback are supposed to be done using a simple execute()

The code should be changed as follows:

pool.acquire(function(err, conn) {
  if (err)
    return sendError(res, err, 500);

  var q = conn.query().insert('FILES', fields, values);
  q.execute(function(err, result) {

    if (err || result.affected < 1 ) {
       pool.release(conn);
       return sendError(res, err, 500);
    }

    conn.query().execute("commit", function(err,result) {
      if (err) {
        pool.release(conn);
        return sendError(res, err, 500);
      }
      res.end('<ok />');
      pool.release(conn);
    });
  });
});

Answer 2

This is not an exact answer to your question, but have a look at node-oracle package. It lacks connection pooling, but its commit / rollback functionality can at least be controlled from code. And you can always mix it with a generic pool solution such as node-pool.