Rackspace Cloud Queues Concepts#
To use the Cloud Queues API effectively, you should understand the key concepts described in this section.
How Cloud Queues works#
Following is an overview of Cloud Queues works:
Create a queue to which producers or publishers post messages.
Workers (consumers or subscribers) claim or get a message from the queue, complete the work in that message, and delete the message.
If a worker will be offline before it completes the work in a message, the worker can retire the claim's time to live (TTL), putting the message back into the queue for another worker to claim.
Subscribers monitor the claims from these queues to track activity and help troubleshoot errors.
For the majority of use cases, Cloud Queues is not responsible for the ordering of messages. However, if there is only a single producer, Cloud Queues ensures that messages are handled in a First In, First Out (FIFO) order.
Cloud Queues supports a variety of messaging patterns including the following:
The producer-consumer pattern has the following characteristics:
- A producer is programmed to send messages to a queue.
- Multiple workers (or consumers) are programmed to monitor a queue.
- Only one worker can claim a message so that no other worker can claim the message and duplicate the work.
- The worker must delete the message when work is done.
- TTL restores a message to an unclaimed state if the worker never finishes.
This pattern is ideal for dispatching jobs to multiple processors.
Characteristics of the publish-subscribe pattern are:
- The publisher sends a message to a queue.
- All workers (or subscribers) listen to the messages in the queue.
- Multiple workers can take action on a message.
- Workers can send a marker or cursor to skip messages already seen.
- TTL eventually deletes messages.
This pattern is ideal for notification of events to multiple workers at once.
Queuing systems are used to coordinate tasks within an application. Here are some examples:
- Backup: A backup application might use a queuing system to connect the actions that users do in the a control panel to the customer's backup agent on a server. When a customer wants to start a backup, they simply choose "start backup" on a panel. Doing so causes the producer to put a "startBackup" message into the queue. Every few minutes, the agent on the customers server (the worker) checks the queue to see if it has any new messages to act on. The agent claims the "startBackup" message and kicks off the backup on the customer's server.
- Storage: Gathering statistics for a large, distributed storage system can be a long process. The storage system can use a queuing system to ensure that jobs complete, even if one initially fails. Since messages are not deleted until after the worker has completed the job, the storage system can make sure that no job goes undone. If the worker fails to complete the job, the message stays in the queue to be completed by another server. In this case, a worker claims a message to perform a statistics job, but the claim's TTL expired and the message is put back into the queue when the job took too long to complete (meaning that it most likely failed). By giving the claim a TTL, applications can protect themselves from workers going offline while processing a message. After a claim's TTL expires, the message is put back into the queue for another worker to claim.
- Email: The team for an email application is constantly migrating customer email from old versions to newer ones, so they develop a tool to let customers do it themselves. The migrations take a long time, so they cannot be done with single API calls, or by a single server. When a user starts a migration job from their portal, the migration tool sends messages to the queue with details of how to run the migration. A set of migration engines, the consumers in this case, periodically check the queues for new migration tasks, claim the messages, perform the migration, and update a database with the migration details. This process allows a set of servers to work together to accomplish large migrations in a timely manner.
Following are some generic use cases for Cloud Queues:
- Distribute tasks among multiple workers (transactional job queues)
- Forward events to data collectors (transactional event queues)
- Publish events to any number of subscribers (publish-subscribe)
- Send commands to one or more agents (point-to-point or publish-subscribe)
- Request an action or get information from a Remote Procedure Call (RPC) agent