Hi, I had this question, I created a cluster with two nodes and the keyspace with these replicas “{‘class’: ‘NetworkTopologyStrategy’, ‘replication_factor’ : 2} AND durable_writes = true;”. Doing an ab test with 500,000 connections, the ab test doesn’t give me any errors but if I do a count inside the table I find less and less data, almost half. Could anyone give me an explanation?
What do you mean by “ab” test? Do you have another database with the same dataset and you execute reads against both of them, comparing results?
Which version of ScyllaDB is this?
ab stands for apache-bench which is a tool for testing concurrent users and connections. The version of scylla is 5.1
What do you mean by “less and less”? Does a count query return different results on each invocation?
Could you please try writing a script which does a full scan of the table and manually counts the rows? See if that has a correct result?
I insert 500,000 rows with the test but doing a count on the db gives me much less, as in the screenshots
Can you please try the experiment I asked about above, and report the result here?
Curious, I wonder if it was inserting unique rows or duplicate rows, and what was the schema.