Databases
We'll not recommend which database to use but instead share our experience with the Node.js clients used to connect to whichever database your project/organization has chosen.
In most cases there is only a single client with widespread usage, often provided by the project or organizations which builds the database itself.
Recommended packages#
Members of the team have had success using the following clients:
| Server | Recommended client |
|---|---|
| Cloudant | @ibm-cloud/cloudant |
| CouchDB | nano |
| Elasticsearch | @elastic/elasticsearch |
| Generic SQL | odbc |
| IBM Db2 | ibm_db, odbc |
| Mongo | mongodb |
| Postgress | pg |
| Redis | ioredis |
There are popular databases for which there are no recommendations. However, if the team did not have some experience with the Node.js client we've not included it in the list.
General Guidance#
Many clients support connection pooling. When available use pooling to avoid creating a new connection for each request.
By default many clients will return all results for a query unless you take steps to limit the result size. Unless you are sure the result set will always be a manageable size use avaliable options to return only a subset of results at a time.
Some of the the clients require a native database binding to be installed. This may limit the platforms on which you can run the client. Some may also support both pure JavaScript and native bindings. In those cases pure JavaScript will be easier to install but the native binding may provide better performance.
Postgress#
Recommendation is to use pg as the most widely used (if not only) client.
Mongo#
Recommendation is to use mongodb which is the official MongoDB driver provided by the Mongo project.
Redis#
Recommendation is to use ioredis. There are 2 good options with no clear technical winner, but IBM has experience in production with ioredis so that is why it is recommended over redis.
Guidance#
If you are already using redis for messaging this can provide a low cost option but does not support more advanced features like High Availability.
Cloudant#
Current versions of cloundant require a cloudant specific client which is [cloudant-node-sdk][]. While it is version 0.0.16 as when this section was last updated, it is a fork of the nano CouchDB client which has a long history. This client can also be used to access CouchDB instances if you are using both Cloudant and CouchDB.
If you are using classic Cloudant instances then you can use the clients recommended for CouchDB
IBM Db2#
On databases except for Db2 for i the recommendation is to use ibm_db which is the official driver provided by IBM.
For the Db2 for i database the primary recommendation is to use odbc which is maintained by IBM and allows you to do local development and then deploy your application to the IBM i platform. If you are doing both development and depolyment on IBM i then using idb-connector, idb-pconnector are good alternatives which require no configuration or additional installation.
Guidance#
- On IBM i clients can inherit authentication based on the Operating system user. In these cases ensure you run the Node.js program using the client as non-admin user.
Elasticsearch#
Recommendation is to use @elastic/elasticsearch which is the official driver provided by Elastic. Note that this client is replacing the original client provided by Elastic which is being deprecated.
CouchDB#
The primary recommendation is to use nano which is the official Apache CouchDB client.
A secondary recommendation is pouchdb which makes it easy to have a local copy which is kept in sync and is great for unit testing.
Generic SQL#
If there is no database specific client or you want your code to be database agnostic the odbc client is a good choice. It's currently maintained by IBM team and is the recommended client for access to the IBM i database.
This client should work for any database for which there is an odbc compliant driver.
Guidance#
Use the built in odbc connection pooling support to avoid creating/destroying a connecttion for each request.
Use the odbc built in functions for:
- isolation/commit
- transaction support
instead of the SQL equivalents.