Collecting slow search queries

Steps to help you identify slow search queries using the DataStax Enterprise Performance Service.

The solr_slow_sub_query_log_options performance object reports distributed sub-queries (query executions on individual shards) that take longer than a specified period of time.

All objects are disabled by default.


The location of the dse.yaml file depends on the type of installation:
Package installations /etc/dse/dse.yaml
Tarball installations installation_location/resources/dse/conf/dse.yaml


To identify slow search queries using the DataStax Enterprise Performance Service:
  1. To permanently enable and configure collecting information on slow search queries, edit the dse.yaml file and uncomment the solr_slow_sub_query_log_options parameters, and define values for the CQL slow log settings:
        ttl_seconds: 604800
        async_writers: 1
        threshold_ms: 100
    The default parameter values minimize resource usage.
    Options Determines
    enabled Whether the object is enabled at start up.
    ttl_seconds How many seconds a record survives before it is expired from the performance object.
    async_writers For event-driven objects, such as the slow log, determines the number of possible concurrent slow query recordings. Objects like solr_result_cache_stats are updated in the background.
    threshold_ms For the slow log, the level (in milliseconds) at which a sub-query slow enough to be reported.
  2. To temporarily change the running parameters for collecting information on slow Solr queries:

    To temporarily enable collecting information:

    dsetool perf solrslowlog enable
    To temporarily disable collecting information:
    dsetool perf solrslowlog disable
    To temporarily change the threshold value in milliseconds:
    dsetool perf solrslowlog 200
  3. You can export slow search queries using the CQL command:
    cqlsh:dse_perf> COPY solr_slow_sub_query_log ( date, commands, duration ) TO 'slow_solr_queries.csv' WITH HEADER = true;