InfluxQL query management
This page documents an earlier version of InfluxDB OSS. InfluxDB OSS v2 is the latest stable version.
Manage your InfluxQL queries using the following:
- SHOW QUERIES to identify currently-running queries
- KILL QUERIES to stop queries overloading your system
- Configuration settings to prevent and halt the execution of inefficient queries
The commands and configurations provided on this page are for Influx Query Language (InfluxQL) only – no equivalent set of Flux commands and configurations currently exists. For the most current Flux documentation, see Get started with Flux.
List currently-running queries with SHOW QUERIES
SHOW QUERIES
lists the query id, query text, relevant database, and duration
of all currently-running queries on your InfluxDB instance.
Syntax
SHOW QUERIES
Example
> SHOW QUERIES
qid query database duration status
--- ----- -------- -------- ------
37 SHOW QUERIES 100368u running
36 SELECT mean(myfield) FROM mymeas mydb 3s running
Explanation of the output
qid
: The id number of the query. Use this value withKILL - QUERY
.query
: The query text.database
: The database targeted by the query.duration
: The length of time that the query has been running. See Query Language Reference for an explanation of time units in InfluxDB databases.SHOW QUERIES
may output a killed query and continue to increment its duration until the query record is cleared from memory.status
: The current status of the query.
Stop currently-running queries with KILL QUERY
KILL QUERY
tells InfluxDB to stop running the relevant query.
Syntax
Where qid
is the query ID, displayed in the SHOW QUERIES
output:
KILL QUERY <qid>
InfluxDB Enterprise clusters: To kill queries on a cluster, you need to specify the query ID (qid) and the TCP host (for example, myhost:8088
),
available in the SHOW QUERIES
output.
KILL QUERY <qid> ON "<host>"
A successful KILL QUERY
query returns no results.
Examples
-- kill query with qid of 36 on the local host
> KILL QUERY 36
>
-- kill query on InfluxDB Enterprise cluster
> KILL QUERY 53 ON "myhost:8088"
>
Configuration settings for query management
The following configuration settings are in the coordinator section of the configuration file.
max-concurrent-queries
The maximum number of running queries allowed on your instance.
The default setting (0
) allows for an unlimited number of queries.
If you exceed max-concurrent-queries
, InfluxDB does not execute the query and
outputs the following error:
ERR: max concurrent queries reached
query-timeout
The maximum time for which a query can run on your instance before InfluxDB
kills the query.
The default setting ("0"
) allows queries to run with no time restrictions.
This setting is a duration literal.
If your query exceeds the query timeout, InfluxDB kills the query and outputs the following error:
ERR: query timeout reached
log-queries-after
The maximum time a query can run after which InfluxDB logs the query with a
Detected slow query
message.
The default setting ("0"
) will never tell InfluxDB to log the query.
This setting is a duration literal.
Example log output with log-queries-after
set to "1s"
:
[query] 2016/04/28 14:11:31 Detected slow query: SELECT mean(usage_idle) FROM cpu WHERE time >= 0 GROUP BY time(20s) (qid: 3, database: telegraf, threshold: 1s)
qid
is the id number of the query.
Use this value with KILL QUERY
.
The default location for the log output file is /var/log/influxdb/influxdb.log
. However on systems that use systemd (most modern Linux distributions) those logs are output to journalctl
. You should be able to view the InfluxDB logs using the following command: journalctl -u influxdb
max-select-point
The maximum number of points that a
SELECT
statement can process.
The default setting (0
) allows the SELECT
statement to process an unlimited
number of points.
If your query exceeds max-select-point
, InfluxDB kills the query and outputs
the following error:
ERR: max number of points reached
InfluxDB checks the point count every second (so queries exceeding the maximum aren’t immediately aborted).
max-select-series
The maximum number of series that a
SELECT
statement can process.
The default setting (0
) allows the SELECT
statement to process an unlimited
number of series.
If your query exceeds max-select-series
, InfluxDB does not execute the query
and outputs the following error:
ERR: max select series count exceeded: <query_series_count> series
max-select-buckets
The maximum number of GROUP BY time()
buckets that a query can process.
The default setting (0
) allows a query to process an unlimited number of
buckets.
If your query exceeds max-select-buckets
, InfluxDB does not execute the query
and outputs the following error:
ERR: max select bucket count exceeded: <query_bucket_count> buckets
Was this page helpful?
Thank you for your feedback!
Support and feedback
Thank you for being part of our community! We welcome and encourage your feedback and bug reports for InfluxDB and this documentation. To find support, use the following resources:
Customers with an annual or support contract can contact InfluxData Support.