thecubed/prestowatcher

Watch your PrestoDB queries to make sure they're not being silly.

prestowatcher

A very simple Slack alerter for PrestoDB queries that are potentially unintentionally resource intensive.

Criteria

Currently prestowatcher only supports alerting if a given source used in a query is scanning over more than a configurable amount of partitions.

Whitelisting Queries

Simply add -- sqlbandit:off somewhere in your query and it'll ignore this query.

Future

Future features might include checking for missing filters and query runtimes.

Building

git clone https://github.com/thecubed/prestowatcher
docker build prestowatcher

Running

Usage:
  prestowatcher [OPTIONS]

Application Options:
  -v, --verbose   Enable DEBUG logging
  -V, --version   Print version and exit
  -u, --url=      presto URL (including scheme and port) [$PRESTO_URL]
  -m, --maxpart=  Alert when Presto queries scan more than X partitions (default: 30) [$MAX_PARTITIONS]
  -i, --interval= Update interval in seconds (default: 20) [$UPDATE_INTERVAL]
  -s, --slack=    Slack Webhook URL [$SLACK_URL]
  -p, --port=     Health check HTTP server port (default: 8080) [$PORT]

Help Options:
  -h, --help      Show this help message

Any options with a $NAME in the help are able to be specified as environment variables to ease deployment in cloud environments.

The application exposes a HTTP health check at / which will return the last successful time it was able to check queries in Presto. Use this as a health check if running under Marathon/Kubernetes to ensure the service isn't stuck.