Columnar segment queries impacting cluster cpu usage


#1

Hi -

Every 5 minutes, there are two queries that are executing on my cluster (apparently by MemSQL – not by any of our applications):

SELECT
  DATABASE_NAME,
  TABLE_NAME,
  COLUMN_NAME,
  SUM(COMPRESSED_SIZE) AS comp_size,
  SUM(UNCOMPRESSED_SIZE) AS uncomp_size
FROM
  information_schema.COLUMNAR_SEGMENTS
GROUP BY
  DATABASE_NAME,
  TABLE_NAME,
  COLUMN_NAME;
SELECT
  IFNULL(SUM(COMPRESSED_SIZE), 0) AS diskUsage
FROM
  INFORMATION_SCHEMA.COLUMNAR_SEGMENTS;

When these queries run, they take 1m+ to run and cause the CPU on all leaf nodes to spike to close to 100% Admittedly I have many large columnstore tables currently being stored in the cluster, but they’re not actively being written to or read from. Is there any way to reduce the impact of these queries on the cluster?

I’m currently running MemSQL 6.5.8.


#2

These queries are issued by the schema monitoring component of ops. You can disable this by opening the ops ui, going to settings, and unchecking the enabled box under “MemSQL Schema Monitoring”.


#3

Perfect – thank you!