collect serverStatus, Oplog and others metrics without --collect-all

[DBaffaleuf-Tac]

Hi there,

I am testing an exporter in a docker container using --collect-all, but this is consuming way too much CPU on the target, so I am trying to find a way to limit the number of metrics collected by the exporter and enabling only the ones I need :

docker run -d --name "percona-mg-exporter" -p 9216:9216 \
	--expose 9216 percona/mongodb_exporter:0.41.1 \ 
	--mongodb.uri="mongodb://mongodb-percona-mg-exporter:passwd@host:27017/admin?ssl=false" \
	--mongodb.collstats-colls=base.col1,base.col2 \ 
	--collector.pbm \
	--collector.dbstats
	--collector.collstats 

Yet there are couple of them I can't find in the documentation : https://github.com/percona/mongodb_exporter/blob/main/REFERENCE.md

1) Metrics related to db.serverStatus():
mongodb_ss*_

2) Metrics related to oplog stats:
mongodb_oplog_stats*_

3) and this one : mongodb_members_configTerm

Is there any ay to collect these counters without having to collect everything (especially the serverStatus ones) ?
Thanks,

[BupycHuk]

Hi, can you try to enable diagnosticdata collector?

[DBaffaleuf-Tac]

Hi thanks it worked , I now have all the metrics I need :

docker run -d --name "percona-mg-exporter" -p 9216:9216 \ 
--expose 9216 percona/mongodb_exporter:0.41.1 \ 
--mongodb.uri="mongodb://mongodb-percona-mg-exporter:password@host:27017/admin?ssl=false" \
--mongodb.collstats-colls=base1.col1,base1.col2 \
--collector.pbm \
--collector.diagnosticdata \
--collector.dbstats \
--collector.collstats \
--collector.replicasetstatus

However, it seems to be leaking CPU% and MEM% over time. I restarted the container yesterday evening, consuming :

# ps -p $(pidof mongodb_exporter) -o %cpu,%mem,cmd
%CPU %MEM CMD
 0.7  0.2 /mongodb_exporter ...

And this morning :

ps -p $(pidof mongodb_exporter) -o %cpu,%mem,cmd
%CPU %MEM CMD
14.0  8.4 /mongodb_exporter ... 

[DBaffaleuf-Tac]

It looks like connections are leaked. I had 1400+ connections to the node at some point, I restarted the container, now I am back at 11 connections, could this be related to the GO Driver (1.16.1) ?

[BupycHuk]

We fixed connection leak in 0.41.2

[BupycHuk]

We found root cause of the problem and released v0.42.0.