Performance of [ACT_RU_METER_LOG]

Camunda 7 Topics Discussion & Questions
1

We recently experience very high load on a sql server hosting the camunda db.

The culprit turned out to be queries of [ACT_RU_METER_LOG] and a quick fix was creating this index:

CREATE NONCLUSTERED INDEX [themissingindex, dbo>]
ON [dbo].[ACT_RU_METER_LOG] ([NAME_],[TIMESTAMP_])
INCLUDE ([VALUE_])

It seems the root cause was > 600000 rows in this table (combined with the missing index); what is the best practice here to avoid that amount of rows and still keep metrics for recent runs?

2

Only the create scripts for H2 and DB2 database contain this, or at least a very similar, index as well.
For all other databases it’s not created by default.

activiti.db2.create.engine.sql:CREATE INDEX ACT_IDX_METER_LOG ON ACT_RU_METER_LOG(NAME_, TIMESTAMP_);
activiti.h2.create.engine.sql:CREATE INDEX ACT_IDX_METER_LOG ON ACT_RU_METER_LOG(NAME_, TIMESTAMP_);

Why? I don’t know.

1 Like
3

I found the problem. It turns out adding Prometheus monitoring caused the reporting interval for [ACT_RU_METER_LOG] to change from 15 minutes to 5 seconds. As it writes 10 rows / iteration, that caused extreme growth of data combined with frequent scans of the whole index.

For reference if anyone else sees the problem, the plugin is:

4