Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/emicklei/parzello
Parzello is a delay and retry service on top of Google Pub Sub to publish messages with a time delay to a topic.
https://github.com/emicklei/parzello
cloud gcp google messaging pubsub
Last synced: 13 days ago
JSON representation
Parzello is a delay and retry service on top of Google Pub Sub to publish messages with a time delay to a topic.
- Host: GitHub
- URL: https://github.com/emicklei/parzello
- Owner: emicklei
- Created: 2018-03-31T05:23:20.000Z (over 6 years ago)
- Default Branch: master
- Last Pushed: 2024-02-24T14:15:39.000Z (8 months ago)
- Last Synced: 2024-10-04T19:16:17.938Z (about 1 month ago)
- Topics: cloud, gcp, google, messaging, pubsub
- Language: Go
- Homepage:
- Size: 4.19 MB
- Stars: 6
- Watchers: 4
- Forks: 0
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# parzello
Parzello is a delay service on top of Google Pub Sub to publish messages with a time delay to a topic.
It uses multiple intermediate topic->subscriptions to queue such messages which are pulled (streaming) on computed time intervals.
These intermediate resources (topics, subscription) need to be created before deploying this service with the corresponding configuration.## use cases
Parzello was created to support a couple of unusual usecases.
### backoff or postpone processing
Consider the situation that the subscriber is unable to process a message because its backend services are not available.
Each such message can not be acknowledged.
Therefore it will be offered on the next pull (or push) causing many operation calls which cause increased network traffic.
One solution is to deliver those messages to `parzello` for the purpose of retrying it later.
Alternatively, the subscriber could be given the capability to stop|pause the pulling process.### late night or bulk processing
Consider the requirement that messages need to be processed in a certain time window only (at night when more resources are available).
One solution is to deliver those messages to `parzello` for the purpose of having them published later.### out of order
Pub Sub, like most message queueing solutions, does not guarantee that messages are delivered in the same order as they were published. This must be taken into account when designing the message definition and the subscriber software architecture. So rather then publishing events such as "order created", "order detail changed" one should have a single event "order updated" with all information about the order and the timestamp when it was updated.
If this design is not possible then the subscriber must handle the situation that an "order changed" event is received before an "order created" for the same entity.
One solution is to deliver those messages to `parzello` for the purpose of having them published again later.
Ofcourse this is a fragile design and can lead to problems (what if the created event was never published?) so be aware of this when using `parzello` for this case.## how does it work?
Instead of the publisher publishing to its destination topic, it publishes the message to a `parzello` PubSub topic specifying the time to delay the publish using message properties (see details below).
The `parzello` service will publish that message to one of its intermediary topics.
Each such topic has its own listener (pulling messaging) at a time interval.
Using a configuration, you must specify which intermediary topics you have created with durations based on your estimations of the actual delay amounts.![](./doc/parzello_delay.png)
In this diagram, 2 such topics exists. One is pulled every 5 minutes.
Each message is inspected, by looking at the message properties, to see whether it is about time to publish it to the actual destination topic.For example, a message that needs to be delivered after 8 minutes will be published by `parzello` to `parzello_5_minutes` once and 3 times to `parzello_1_minute `.
![](./doc/parzello_delay_retry.png)
In the next design, `parzello` is used to retry publishing a message at a later moment.
By passing publish count metadata (parzello.publishCount) to the message, a subscriber can inspect this value and behave accordingly (i.e abort on when this value exceeds the maximum retries).## mirror messages
When messages are being retried multiple times, it can very helpful to inspect those messages to see the reason and the payload of such a message. To support this in `parzello`, messages can be mirrored in a DataStore by indicating this using the `parzello.datastoreLookup` and `parzello.datastoreInfo` properties. Message stored in DataStore can be easily queried and counted (see REST API) using the Google Cloud Console. Once a message is published to its destination topic, it will also be deleted from the DataStore.
![](./doc/parzello_datastore.png)
( Name,DestinationTopic,EntryTime,Info,Lookup,Payload,Properties,PublishAfter,PublishCount,PublishTime )
## usage
### message attributes
#### properties when publishing to `parzello`
|name |required |comment
|---------------------------|-----------|--------
|parzello.destinationTopic |true |topic to which the message eventually must be published
|parzello.publishAfter |true |Unix time (seconds after 1970) after which the message must be published
|parzello.publishCount |false |if set then the value is incremented before publishing to the destionationTopic
|parzello.datastoreLookup |false |if set to some lookup value (e.g. an entity identifier) then also store the message with payload in DataStore for querying
|parzello.datastoreInfo |false |use this field to add context information to the message stored in the DataStore
|X-Cloud-Debug |false |if set to some identifier then for this message debug logging is produced#### properties on a message received from `parzello`
|name |comment
|---------------------------|-------
|parzello.publishCount |set to 1 if missing otherwise it is incremented when received by parzello. So the publisher should add this property on every retry.## REST API
The REST API is for querying the DataStore using simple queries that may not be possible using GQL, the query language for Google DataStore.
#### /v1/count?Lookup=your-value&Info=other-value
Returns the number of mirrored messages stored in the DataStore.
If the `Lookup` is given then use that as a filter (WHERE condition).
If the `Info` is given then use that as a filter (WHERE condition).## Config
The `parzello` server configuration (YAML) lists one or more queues (intermediate topic+subscriptions).
This example show 2 such queues.
The topics and subscriptions must be created upfront in the project `YOUR-PROJECT-ID`.
A duration string is a possibly signed sequence of decimal numbers, each with optional fraction and a unit suffix, such as "60s", "1.5h" or "2h45m". Valid time units are "ms", "s", "m", "h".project-id: YOUR-PROJECT-ID
subscription: parzello_inbound
queues:
- topic: parzello_minute
subscription: parzello_minute
duration: 1m
- topic: parzello_five_minutes
subscription: parzello_five_minutes
duration: 5m
basic-auth:
username: parzello
password: parzello## Google AppEngine Deployment
### Google PubSub topics and subscriptions
Before deploying the parzello application, you need to setup the required infrastructure:
- make sure Google AppEngine is enabled
- make sure all PubSub topics mentioned in the configuration are created
- make sure all PubSub subscriptions mentioned in the configuration are createdSee [https://cloud.google.com/sdk/gcloud/reference/pubsub/](https://cloud.google.com/sdk/gcloud/reference/pubsub/)
### deploy as a service
As the app.yaml specifies the service to be `parzello`, Google AppEngine needs to have an application deployed as default first.
Alternatively, you modify the app.yaml file and set the service to default instead.
Note that `parzello` requires Flex deployment because it starts background processes to consume subscriptions.gcloud app deploy
## Development
make run