diaspora-event-sdk 0.0.19

SDK of Diaspora Event Fabric: Resilience-enabling services for science from HPC to edge

Diaspora Event Fabric: Resilience-enabling services for science from HPC to edge

• Installation
  • Recommended Installation with Kafka Client Library
  • Installation Without Kafka Client Library
• Use Diaspora Event SDK
  • Use the SDK to communicate with Kafka (kafka-python Required)
    • Register Topic (create topic ACLs)
    • Block Until Ready
    • Start Producer
    • Start Consumer
    • Unregister Topic (remove topic ACLs)
  • Use Your Preferred Kafka Client Library
    • Register and Unregister Topic
    • Cluster Connection Details
  • Advanced Usage
    • Password Refresh
• Common Issues
  • ImportError: cannot import name 'KafkaProducer' from 'diaspora_event_sdk'
  • kafka.errors.NoBrokersAvailable and kafka.errors.NodeNotReadyError
  • kafka.errors.KafkaTimeoutError: KafkaTimeoutError: Failed to update metadata after 60.0 secs.
  • ssl.SSLCertVerificationError

Installation

Recommended Installation with Kafka Client Library

If you plan to utilize the KafkaProducer and KafkaConsumer classes in the SDK, which are extensions of the respective classes from the kafka-python library, we recommend installing the SDK with kafka-python support. This is especially convenient for tutorial purposes and integrating Kafka functionalities in your projects with out-of-box configurations.

To install Diaspora Event SDK with kafka-python, run:

pip install "diaspora-event-sdk[kafka-python]"

Installation Without Kafka Client Library

For scenarios where kafka-python is not required or if you are using other client libraries to communicate with Kafka, you can install the SDK without this dependency.

To install the SDK without Kafka support, simply run:

pip install diaspora-event-sdk

Note that this option does not install the necessary dependency for KafkaProducer and KafkaConsumer below to work.

Use Diaspora Event SDK

Use the SDK to communicate with Kafka (kafka-python Required)

Register Topic (create topic ACLs)

Before you can create, describe, and delete topics we need to set the appropriate ACLs in ZooKeeper. Here we use the Client to register ACLs for the desired topic name.

from diaspora_event_sdk import Client as GlobusClient
c = GlobusClient()
topic = "topic-" + c.subject_openid[-12:]
print(c.register_topic(topic))
print(c.list_topics())

# registration_status = c.register_topic(topic)
# print(registration_status)
# assert registration_status["status"] in ["no-op", "success"]

# registered_topics = c.list_topics()
# print(registered_topics)
# assert topic in registered_topics["topics"]

Register a topic also creates it, if the topic previously does not exist.

Block Until Ready

KafkaProducer and KafkaConsumer would internally call create_key if the the connection credential is not found locally (e.g., when you first authenticated with Globus). Behind the sence, the middle service contacts AWS to initialize the asynchronous process of creating and associating the secret. The method below blocks until the credential is ready to be used by producer and consumer. When the method finishes, it returns True and the producer and consumer code below should work without further waiting. By default, the method retries in loop for five minutes before giving up and return False. Use parameter max_minutes to change the number of minutes of max waiting.

from diaspora_event_sdk import block_until_ready
assert block_until_ready()

Start Producer

Once the topic is created we can publish to it. The KafkaProducer wraps the Python KafkaProducer Event publication can be either synchronous or asynchronous. Below demonstrates the synchronous approach.

from diaspora_event_sdk import KafkaProducer
producer = KafkaProducer()
future = producer.send(
    topic, {'message': 'Synchronous message from Diaspora SDK'})
print(future.get(timeout=10))

Start Consumer

A consumer can be configured to monitor the topic and act on events as they are published. The KafkaConsumer wraps the Python KafkaConsumer. Here we use the auto_offset_reset to consume from the first event published to the topic. Removing this field will have the consumer act only on new events.

from diaspora_event_sdk import KafkaConsumer
consumer = KafkaConsumer(topic, auto_offset_reset='earliest')
for msg in consumer:
    print(msg)

Unregister Topic (remove topic ACLs)

from diaspora_event_sdk import Client as GlobusClient
c = GlobusClient()
topic = "topic-" + c.subject_openid[-12:]
print(c.unregister_topic(topic))
print(c.list_topics())

Use Your Preferred Kafka Client Library

Register and Unregister Topic

The steps are the same as above by using the register_topic, unregister_topic, and list_topics methods from the Client class.

Cluster Connection Details

Configuration	Value
Bootstrap Servers	MSK_SCRAM_ENDPOINT
Security Protocol	SASL_SSL
Sasl Mechanism	SCRAM-SHA-512
Api Version	3.5.1
Username	(See instructions below)
Password	(See instructions below)

Execute the code snippet below to obtain your unique username and password for the Kafka cluster:

from diaspora_event_sdk import Client as GlobusClient
c = GlobusClient()
print(c.retrieve_key())

Advanced Usage

Key Migration

In case you want to use the same credential (OpenID, secret key) on the second macine, but a call to create_key() invalidates all previous keys so you cannot rely on this API for generating a new key for the new machine, while keeping the old key on the old machine valid. Starting at 0.0.19, you can call the following code to retrieve the active secret key on the first machine and store it to the second machine, so that both machines hold valid keys.

On the first machine, call:

from diaspora_event_sdk import Client as GlobusClient
c = GlobusClient()
print(c.get_secret_key()) # note down the secret key

On the second machine, call:

from diaspora_event_sdk import Client as GlobusClient
from diaspora_event_sdk import block_until_ready

c = GlobusClient()
c.put_secret_key("<secret-key-from-first-machine>")
assert block_until_ready() # should unblock immediately

Note that a call to create_key() on any machine would invalidate the current key, on both machines.

Password Refresh

In case that you need to invalidate all previously issued passwords and generate a new one, call the create_key method from the Client class

from diaspora_event_sdk import Client as GlobusClient
c = GlobusClient()
print(c.create_key())

Subsequent calls to retrieve_key will return the new password from the cache. This cache is reset with a logout or a new create_key call.

Common Issues

ImportError: cannot import name 'KafkaProducer' from 'diaspora_event_sdk'

It seems that you ran pip install diaspora-event-sdk to install the Diaspora Event SDK without kafka-python. Run pip install kafka-python to install the necessary dependency for our KafkaProducer and KafkaConsumer classes.

kafka.errors.NoBrokersAvailable and kafka.errors.NodeNotReadyError

These messages might pop up if create_key is called shortly before instanciating a Kafka client. This is because there's a delay for AWS Secret Manager to associate the newly generated credential with MSK. Note that create_key is called internally by kafka_client.py the first time you create one of these clients. Please wait a while (around 1 minute) and retry.

kafka.errors.KafkaTimeoutError: KafkaTimeoutError: Failed to update metadata after 60.0 secs.

Step 1: Verify Topic Creation and Access: Before interacting with the producer/consumer, ensure that the topic has been successfully created and access is granted to you. Execute the following command:

from diaspora_event_sdk import Client as GlobusClient
c = GlobusClient()
# topic = <the topic you want to use>
print(c.register_topic(topic)) 

This should return a status: no-op message, indicating that the topic is already registered and accessible.

Step 2: Wait Automatic Key Creation in KafkaProducer and KafkaConsumer KafkaProducer and KafkaConsumer would internally call create_key if the keys are not found locally (e.g., when you first authenticated with Globus). Behind the sence, the middle service contacts AWS to initialize the asynchronous process of creating and associating the secret. Please wait a while (around 1 minute) and retry.

ssl.SSLCertVerificationError

This is commmon on MacOS system, see this StackOverflow answer.