This plugin is a drop-in replacement of the DefaultJobQueuePlugin, which implements a push-based job queue strategy built on top of the popular BullMQ library.
The advantage of this approach is that jobs are stored in Redis rather than in the database. For more complex applications with many job queues and/or multiple worker instances, this can massively reduce the load on the DB server. The reason for this is that the DefaultJobQueuePlugin uses polling to check for new jobs. By default it will poll every 200ms. A typical Vendure instance uses at least 3 queues (handling emails, collections, search index), so even with a single worker instance this results in 15 queries per second to the DB constantly. Adding more custom queues and multiple worker instances can easily result in 50 or 100 queries per second. At this point performance may be impacted.
Using this plugin, no polling is needed, as BullMQ will push jobs to the worker(s) as and when they are added to the queue. This results in significantly more scalable performance characteristics, as well as lower latency in processing jobs.
Note: To use this plugin, you need to manually install the bullmq package:
Note: The v1.x version of this plugin is designed to work with bullmq v1.x, etc.
Example
To develop with this plugin, you'll need an instance of Redis to connect to. Here's a docker-compose config that will set up Redis as well as Redis Commander, which is a web-based UI for interacting with Redis:
The default concurrency of a single worker is 3, i.e. up to 3 jobs will be processed at the same time.
You can change the concurrency in the workerOptions passed to the init() method:
Example
By default, BullMQ will keep completed jobs in the completed set and failed jobs in the failed set. Over time,
these sets can grow very large. Since Vendure v2.1, the default behaviour is to remove jobs from these sets after
30 days or after a maximum of 5,000 completed or failed jobs.
This can be configured using the removeOnComplete and removeOnFail options:
Example
The count option specifies the maximum number of jobs to keep in the set, while the age option specifies the
maximum age of a job in seconds. If both options are specified, then the jobs kept will be the ones that satisfy
both properties.
Some jobs are more important than others. For example, sending out a timely email after a customer places an order is probably more important than a routine data import task. Sometimes you can get the situation where lower-priority jobs are blocking higher-priority jobs.
Let's say you have a data import job that runs periodically and takes a long time to complete. If you have a high-priority job that needs to be processed quickly, it could be stuck behind the data import job in the queue. A customer might not get their confirmation email for 30 minutes while that data import job is processed!
To solve this problem, you can set the priority option on a job. Jobs with a higher priority will be processed before
jobs with a lower priority. By default, all jobs have a priority of 0 (which is the highest).
Learn more about how priority works in BullMQ in their documentation.
You can set the priority by using the setJobOptions option (introduced in Vendure v3.2.0):
Example
By default, the underlying BullMQ library will use the default Redis key prefix of bull. This can be changed by setting the prefix option
in the queueOptions and workerOptions objects:
BullMQPluginOptions(options: BullMQPluginOptions) => Configures the plugin.
Architecture reviews, custom plugin work, migrations, ongoing support. Get a hand from the team that builds Vendure.
Talk to the team