Additional Prometheus metrics in BKG NtripCaster

[anonymous]

Hi BKG team,

There is a nice feature that you added in V2.0.42, that is the provision of Prometheus metrics, that we are using a lot for monitoring our BKG NtripCasters with Grafana, and we think that it would be super helpful having additional metrics to complement the current ones, in this case per mountpoint, for having a complete monitoring of the casters in real-time, like:

• Received bytes per mountpoint.
• Sent bytes per mountpoint.
• Number of clients per mountpoint
• Number of sources per mountpoint (this would be always 1 or 0 but it would help to know the current active mountpoints)
• Last time traffic was received per mountpoint.
• Last time traffic was sent per mountpoint.
• Received bytes per relay.
• Last time traffic was received per relay.
• Errors per relay and last time there was an error per relay?

Thank you in advance!
Jorge

[stoecker]

In 10501/ntrip:

see #183, first stream statistics

[stoecker]

In 10502/ntrip:

see #183, small cleanup

[stoecker]

In 10504/ntrip:

see #183, move load, increase version

[wiese]

added:

# HELP sources_received_bytes_total The number of bytes received for the mountpoint.
# TYPE sources_received_bytes_total counter
sources_received_bytes_total{mp="FFMJ"} 3586048

# HELP sources_sent_bytes_total The number of bytes sent for the mountpoint.
# TYPE sources_sent_bytes_total counter
sources_sent_bytes_total{mp="FFMJ"} 27648

# HELP sources_clients_num The number of clients connected to the mountpoint.
# TYPE sources_clients_num gauge
sources_clients_num{mp="FFMJ"} 1

# HELP sources_connections_total The number of connections of the mountpoint.
# TYPE sources_connections_total counter
sources_connections_total{mp="FFMJ"} 1

# HELP sources_clients_connections_total The number of client connections to the mountpoint.
# TYPE sources_clients_connections_total counter
sources_clients_connections_total{mp="FFMJ"} 1

# HELP sources_duration_seconds The activity time of the mountpoint.
# TYPE sources_duration_seconds gauge
sources_duration_seconds{mp="FFMJ"} 4858

[jrocamora@…]

Hi!

Thank you very much for the rapid implementation of this enhancement.

It looks very nice!!

Just a few minor comments:

• I would find useful a metric "sources_connections_num" similar to "sources_clients_num" but with the number of active connections per mountpoint.

• Regarding the "sources_clients_connections_total" metric, would it be computed per mountpoint and client or what is the different between this metric and the "source_connections_total" one?

• Would it be possible having a metric formatted as UNIX timestamp like "sources_last_received_timestamp_seconds" and, optionally, "sources_last_sent_timestamp_seconds"? So that, one is able to monitor the last time a certain mountpoint was receiving/transmitting data. Nevertheless, I don't know if the computation of this metrics would be expensive in terms of CPU as it would be computed very frequently.

• Would the "source_duration_seconds" metric be reset to zero after the connection of a NtripServer is lost or what is the expected behaviour?

Again thank you very much, as it will help integrating and monitoring the BKG NtripCaster in cloud-like environments.

Best regards,
Jorge

[wiese]

Replying to jrocamora@…:

Hi!

Thank you very much for the rapid implementation of this enhancement.

It looks very nice!!

Just a few minor comments:

• I would find useful a metric "sources_connections_num" similar to "sources_clients_num" but with the number of active connections per mountpoint.

  This is sources_clients_num{mp="FFMJ"} 1
  

• Regarding the "sources_clients_connections_total" metric, would it be computed per mountpoint and client or what is the different between this metric and the "source_connections_total" one?

  see examples above
  

• Would it be possible having a metric formatted as UNIX timestamp like "sources_last_received_timestamp_seconds" and, optionally, "sources_last_sent_timestamp_seconds"? So that, one is able to monitor the last time a certain mountpoint was receiving/transmitting data. Nevertheless, I don't know if the computation of this metrics would be expensive in terms of CPU as it would be computed very frequently.

  This is analog to "Connected for" in admin>sources view.
  

• Would the "source_duration_seconds" metric be reset to zero after the connection of a NtripServer is lost or what is the expected behaviour?

  yes it is reset when connection is broken, per mountpoint. See example above
  

Again thank you very much, as it will help integrating and monitoring the BKG NtripCaster in cloud-like environments.

Best regards,
Jorge

[anonymous]

Thank you for the replies! I finally reply to the first questionwiese:

Replying to jrocamora@…:

Hi!

Thank you very much for the rapid implementation of this enhancement.

It looks very nice!!

Just a few minor comments:

• I would find useful a metric "sources_connections_num" similar to "sources_clients_num" but with the number of active connections per mountpoint.

  This is sources_clients_num{mp="FFMJ"} 1
  

  So I understand it is not possible having also a gauge with current connections per mountpoint like "sources_connection_num".
  

• Regarding the "sources_clients_connections_total" metric, would it be computed per mountpoint and client or what is the different between this metric and the "source_connections_total" one?

  see examples above
  

• Would it be possible having a metric formatted as UNIX timestamp like "sources_last_received_timestamp_seconds" and, optionally, "sources_last_sent_timestamp_seconds"? So that, one is able to monitor the last time a certain mountpoint was receiving/transmitting data. Nevertheless, I don't know if the computation of this metrics would be expensive in terms of CPU as it would be computed very frequently.

  This is analog to "Connected for" in admin>sources view.
  

• Would the "source_duration_seconds" metric be reset to zero after the connection of a NtripServer is lost or what is the expected behaviour?

  yes it is reset when connection is broken, per mountpoint. See example above
  

Again thank you very much, as it will help integrating and monitoring the BKG NtripCaster in cloud-like environments.

Best regards,
Jorge

[wiese]

Replying to anonymous:

Thank you for the replies! I finally reply to the first questionwiese:

Replying to jrocamora@…:

Hi!

Thank you very much for the rapid implementation of this enhancement.

It looks very nice!!

Just a few minor comments:

• I would find useful a metric "sources_connections_num" similar to "sources_clients_num" but with the number of active connections per mountpoint.

  This is sources_clients_num{mp="FFMJ"} 1
  

  So I understand it is not possible having also a gauge with current connections per mountpoint like "sources_connection_num".
  

  sources_clients_num is gauge. Please try it out.
  

• Regarding the "sources_clients_connections_total" metric, would it be computed per mountpoint and client or what is the different between this metric and the "source_connections_total" one?

  see examples above
  

• Would it be possible having a metric formatted as UNIX timestamp like "sources_last_received_timestamp_seconds" and, optionally, "sources_last_sent_timestamp_seconds"? So that, one is able to monitor the last time a certain mountpoint was receiving/transmitting data. Nevertheless, I don't know if the computation of this metrics would be expensive in terms of CPU as it would be computed very frequently.

  This is analog to "Connected for" in admin>sources view.
  

• Would the "source_duration_seconds" metric be reset to zero after the connection of a NtripServer is lost or what is the expected behaviour?

  yes it is reset when connection is broken, per mountpoint. See example above
  

Again thank you very much, as it will help integrating and monitoring the BKG NtripCaster in cloud-like environments.

Best regards,
Jorge