Node specified with setNode() method forwards read request to other nodes

Question

I am trying to read data present in a specific cassandra (4) node, using Datastax driver version 4.13. I have read this post, which supposedly works for driver versions 3.x

I have modified my code as best I can for driver version 4.x. The operative part of the code looks like

BoundStatement bs= psGetAllOptimizedModule.bind(imei,year,month,startDate,endDate)
            .setTracing(true)
            .setNode(localNode).setConsistencyLevel(ConsistencyLevel.LOCAL_ONE).setPageSize(pageSize);
    Instant start=Instant.now();
    ResultSet rs = session.execute(bs);

While the query is indeed directed to the specified node, the node then is sometimes forwarding requests to other nodes, even though the data is present on the specified node.

I wanted to check if this is at all possible in cassandra, and if so, what I need to do differently.

Update: As requested, the information asked for has been added to a pastebin here, as its too long to add to the question.

Erick Ramirez · Accepted Answer · 2023-09-13 05:23:49Z

1

The Statement.setNode() API does in fact restrict the request such that the driver will ignore the load balancing policy and only execute the statement on the specified node.

However, the symptom you described indicates to me that the query cannot be satisfied by just one node so it has to request data from other replicas despite setting the consistency to LOCAL_ONE.

In fact the name psGetAllOptimizedModule suggests to me that you are trying to retrieve more than just one record so it makes sense that the coordinating node needs to request the data from other nodes.

You can easily confirm that this is the case by running the query locally on the Cassandra node with tracing enabled and consistency set to LOCAL_ONE. Cheers!

answered Sep 13, 2023 at 5:23

Erick Ramirez

16.5k2 gold badges22 silver badges33 bronze badges

Sign up to request clarification or add additional context in comments.

9 Comments

ranban282 Over a year ago

I am indeed trying to get multiple records from a single partition. The table is incrementally repaired every day. That aside,tracing is enabled, and querying node1 is sometimes answered by node 1, sometimes by node 2 & 3. Similarly, querying node2 results in querying node 2 sometimes, but also node 1 & node 3. Isn't local_one supposed to prevent cross-node queries? Even if the query can't be satisfied by a particular node, it should return an empty set - thats my desired behaviour. As mentioned in the qestion, the only effect of setNode() is to always initially query the desired node.

ranban282 Over a year ago

I have tried to run the query from the cqlsh terminal by turning tracing on, and setting consistency level to local_one. But the session still sometimes hits other nodes, and sometimes it doesn't.

ranban282 Over a year ago

One other question - will changing the load balancing policy help? I'm finding it difficult to find documentation/examples which tells me how I can change/set Load Balancign Policy in DataStax 4.x.

Erick Ramirez Over a year ago

"But the session still sometimes hits other nodes, and sometimes it doesn't." Like I said, the coordinator has to query other nodes other than "local" because the requested data isn't available on the local node. Cheers!

ranban282 Over a year ago

How can the exact same data sometimes be available from the local node, and sometimes not? I am happy to share the trace logs.

|

Collectives™ on Stack Overflow

Node specified with setNode() method forwards read request to other nodes

1 Answer 1

9 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

9 Comments

Your Answer

Sign up or log in

Post as a guest

Related