datastax / astra-sdk-java Goto Github PK

In version 0.2.4 and 0.2.5 we used the jitpack version.

• Configuration of the HTTP client
• Configuration of the Cqlsession with a .conf file

astra.http.connectTimeout=3
astra.http.readtimeout=3
astra.http.keep-alive=3

The current implementation expects the developers to provide BEAN to be populated.

As the unmarshalling / deserialization could be customize allow to work with Raw Json or provide options to implement external mapper.

The application does not end properly when executed in a Main.

On 9/14 the multi region support has been introduced to Astra. On each region (datacenters) you will have:

• A dedicated REST Endpoint
• A dedicated graphQL endpoint
• A dedicated secure connect bundle

As a consequence the drivers cannot failover on their own, they need the secure bundle. The fail-over has to be handle at the SDK level.

AstraClient client = AstraClient.builder()
 .databaseId("astra_cluster_id")  
 .databaseRegion("eu-central-1")
 .databaseFallbackRegions("us-east-1", "us-east-2")
 .keyspace("ks1")
 .token("AstraCS:......") 
 .clientId("TWRvjlcrgfZYfhcxGZhUlAAA") 
 .clientSecret("7xKSrZPLbWxDJ0WXyj..") 
 .build();

We could try add settings on the retry mechanism:

astra.retries.count: 3
astra.retries.healthcheckPeriod: 1min 

Idea:
Register your listener on get notify for each call (audit, monitoring)

• Provide the listener system
• Provide a AuditLogListener
• Provide a PrometheusListener
• Provide a Grafana Dashboard

For a quickstart create a maven archetype to have a read-to-go project

Update documentation to 0.4 and move to Awesome Astra

In application.yaml you can define an extra set of properties

astra:
  cassandra-driver:
    basic.load-balancing-policy:
    local-datacenter: Cassandra 

After the curl install, creating a new Terminal prints out: "/Users//.astra/cli/astra-init.sh:1155: command not found: complete"

When running into Kubernetes Stargate will have its own service doing the load balancing for us.
When running in Astra, a load balancer is already in place.

When running Stargate on your own Cassandra installation you may leverage on a DNS load balancer or use the client side load balancing here.

• Add a way to setup multiple nodes and split per DC.

StargateClient sc = StargateClient.builder()
                // CQL 
                .withAuthCredentials("cassandra","cassandra")
                .withLocalDatacenter("datacenter1")
                .withContactPoints("127.0.0.1:9045")
                .withKeyspace("ks1")
                // API
                .withApiNode(new StargateNode("localhost"))
                .withApiNode(new StargateNode("127.0.0.1"))
                .withApiNodeDC("datacenter1", 
                        new StargateNode("node11",  "http://localhost:8082",  "http://localhost:8080",  "http://localhost:8081"))
                .build()) {

• Implement Load balancing

On some systems (but not all; seems to depend on Maven repository configuration locally), build fails with:

[INFO] BUILD FAILURE
[INFO] ------------------------------------------------------------------------
[INFO] Total time:  11.278 s
[INFO] Finished at: 2022-09-28T10:17:38-07:00
[INFO] ------------------------------------------------------------------------
[ERROR] Failed to execute goal on project stargate-sdk: Could not resolve dependencies for project com.datastax.stargate:stargate-sdk:jar:0.3.5-SNAPSHOT: Failed to collect dependencies at io.grpc:grpc-netty-shaded:jar:1.49.1 -> io.grpc:grpc-core:jar:[1.49.1]: No versions available for io.grpc:grpc-core:jar:[1.49.1] within specified range -> [Help 1]
[ERROR] 

which is somehow triggered by (rare) use of version-range dependency in grpc-netty-shaded:

https://mvnrepository.com/artifact/io.grpc/grpc-netty-shaded/1.49.1

This can be worked around by forcing dependencies at main level.
I will provide a PR for that.

On the free tier, after 23H , Astra will put the database instance in hibernation state.

From there any call will wake up the database. The first call will return 503 coding for a Service not (yet) available. This make the SDK initialization failing.

Enhancement: At initialization, if the code return is 503 corresponding to an hibernated state (need to be confirmed with explict message, the SDK will log it and wait and retry until the DB is available to keep initializing.

Spring boot starter need some breaking changes for the support in SB3

awspring/spring-cloud-aws@da6b929#diff-85078dbef01645df0addf2c549d0a03ebfe2c039296b27980fd9dd65cbd8b94d

Hi Team,
Astra is using DSE which is based on TPC, a lower value of the local connection pool will cause more issues that benefit, can we bump this value to a higher say 8, and add it clearly in documentation to configure it correctly?

    { "name": "astra.cql.driver-config.advanced.connection.pool.local.size",
      "type": "java.lang.Integer",
      "description":"The number of connections in the pool.",
      "defaultValue": "1" },

Proposed a security user details service when security is detected in classpath