question

muzimilbasha_178465 avatar image
muzimilbasha_178465 asked Erick Ramirez edited

Frequent GC flushes leading to connection failures and slow queries

I have simple table like below

CREATE TABLE mytable (
    id text,
    type text,
    value text,
    active boolean,
    PRIMARY KEY (id)
);

The column value can have data ranging from 100kb to 2MB.

I am writing data and reading data into this table frequently. I am experiencing these issues

  1. Long running queries even for simple queries like select active from mytable where id ='1234';
  2. I have also seen frequent GC flushes every 10 to 15 mins with GC pause as ranging from 700 ms to 1 sec
  3. Cassandra connection exceptions

I am using spring data cassandra.

I do not have any tombstones in this table, But the other tables have tombstones.

any suggestions to fix this would be of great help.

cassandra
1 comment
10 |1000

Up to 8 attachments (including images) can be used with a maximum of 1.0 MiB each and 10.0 MiB total.

pmcfadin avatar image pmcfadin ♦♦ commented ·

Are you running this instance on spinning disk or SSD?

0 Likes 0 ·

1 Answer

Erick Ramirez avatar image
Erick Ramirez answered muzimilbasha_178465 commented

@muzimilbasha_178465 the partition size range you quoted sound reasonable but the symptoms you described indicate that the nodes are overloaded and cannot cope.

You haven't provided information on your cluster configuration but I would recommend you first validate your hardware. For pure Cassandra workloads in production, we recommend servers with at least 32GB of RAM and 8 CPU cores with a minimum of 3 nodes per DC. If your use case requires more throughout, then scale out (add more nodes) your cluster.

For details on our hardware recommendations, see Capacity planning and hardware selection for DataStax Enterprise implementations. Cheers!

1 comment Share
10 |1000

Up to 8 attachments (including images) can be used with a maximum of 1.0 MiB each and 10.0 MiB total.

muzimilbasha_178465 avatar image muzimilbasha_178465 commented ·

@Erick Ramirez Thank you for the response. I will check on the infra side.

0 Likes 0 ·