Skip to content

Logql

LogQL is Grafana Loki’s PromQL-inspired query language. Queries act as if they are a distributed grep to aggregate log sources. LogQL uses labels and operators for filtering.

There are two types of LogQL queries:

  • Log queries: Return the contents of log lines.
  • Metric queries: Extend log queries to calculate values based on query results.

Usage⚑

Apply a pattern to the value of a label⚑

Some logs are sent in json and then one of their fields can contain other structured data. You may want to use that structured data to further filter the logs.

{app="ingress-nginx"} | json | line_format `{{.log}}` | pattern `<_> - - <_> "<method> <_> <_>" <status> <_> <_> "<_>" <_>` | method != `GET`
  • {app="ingress-nginx"}: Show only the logs of the ingress-nginx.
  • | json: Interpret the line as a json.
  • ``| line_format}| pattern<> - - <> " <> <>" <> <> "<>" <>```: interpret thelog` json field of the trace with the selected pattern
  • ``| method !=GET````: Filter the line using a key extracted by the pattern.

Count the unique values of a label⚑

Sometimes you want to alert on the values of a log. For example if you want to make sure that you're receiving the logs from more than 20 hosts (otherwise something is wrong). Assuming that your logs attach a host label you can run

sum(count by(host) (rate({host=~".+"} [24h])) > bool 0)

This query will: - {host=~".+"}: Fetch all log lines that contain the label host - count by(host) (rate({host=~".+"} [24h]): Calculates the number of entries in the last 24h. - count by(host) (rate({host=~".+"} [24h])) > bool 0: Converts to 1 all the vector elements that have more than 1 message. - sum(count by(host) (rate({host=~".+"} [24h])) > bool 0): Sums all the vector elements to get the number of hosts that have more than one message.

journald promtail parser is known to fail between upgrades, it's useful too to make an alert to make sure that all your hosts are sending the traces. You can do it with: sum(count by(host) (rate({job="systemd-journal"} [24h])) > bool 0)

Compare the values of a metric with the past⚑

The offset modifier allows changing the time offset for individual range vectors in a query.

For example, the following expression counts all the logs within the last ten minutes to five minutes rather than last five minutes for the MySQL job. Note that the offset modifier always needs to follow the range vector selector immediately.

count_over_time({job="mysql"}[5m] offset 5m) // GOOD
count_over_time({job="mysql"}[5m]) offset 5m // INVALID

References⚑