Metrics of type Summary has been dropped

[paulojmdias]

Hi,

We are using MGOB (https://github.com/stefanprodan/mgob), and MGOB exposes metrics in Prometheus format natively. When we scrape directly from MGOB, the metrics mgob_scheduler_backup_latency (Type summary) are scrapped successfully, but when we use inputs.prometheus, these metrics are probably dropped.

We have metric_version = 2 in inputs.prometheus and outputs.prometheus. If we use metric_version = 1 in inputs.prometheus and outputs.prometheus, this issue won't occurs.

Relevant telegraf.conf:

[[inputs.prometheus]]
  urls = ["http://localhost:8090/metrics"]
  metric_version = 2
  [inputs.prometheus.tags]
      cluster_name = "mongodb-cluster"

System info:

• Telegraf v1.13.2
• Debian 9

Steps to reproduce:

1. Get info from MGOB Exporter

curl -s http://localhost:8090/metrics | grep 'mgob_'

# HELP mgob_scheduler_backup_latency Backup duration in seconds.
# TYPE mgob_scheduler_backup_latency summary
mgob_scheduler_backup_latency{plan="mongodb-cluster",status="200",quantile="0.5"} NaN
mgob_scheduler_backup_latency{plan="mongodb-cluster",status="200",quantile="0.9"} NaN
mgob_scheduler_backup_latency{plan="mongodb-cluster",status="200",quantile="0.99"} NaN
mgob_scheduler_backup_latency_sum{plan="mongodb-cluster",status="200"} 4004.679645687
mgob_scheduler_backup_latency_count{plan="mongodb-cluster",status="200"} 1
# HELP mgob_scheduler_backup_size The size of backup.
# TYPE mgob_scheduler_backup_size gauge
mgob_scheduler_backup_size{plan="mongodb-cluster",status="200"} 1.9250215041e+10
# HELP mgob_scheduler_backup_total The total number of backups.
# TYPE mgob_scheduler_backup_total counter
mgob_scheduler_backup_total{plan="mongodb-cluster",status="200"} 1

2. Get scrape from Telegraf

curl -s http://localhost:9273/metrics | grep 'mgob_'

# HELP mgob_scheduler_backup_size Telegraf collected metric
# TYPE mgob_scheduler_backup_size gauge
mgob_scheduler_backup_size{cluster_name="mongodb-cluster",plan="mongodb-cluster",status="200",url="http://localhost:8090/metrics"} 1.9250215041e+10
# HELP mgob_scheduler_backup_total Telegraf collected metric
# TYPE mgob_scheduler_backup_total counter
mgob_scheduler_backup_total{cluster_name="mongodb-cluster",plan="mongodb-cluster",status="200",url="http://localhost:8090/metrics"} 1

Expected behavior:

Have the same metrics in both requests.

Actual behavior:

Metrics mgob_scheduler_backup_latency (Type summary) has been dropped by Telegraf if we use metric_version = 2.

Additional info:

The summary vector from MGOB: https://github.com/stefanprodan/mgob/blob/master/pkg/metrics/metrics.go#L34