Many metrics are provided by the Collector for its monitoring. Below some
key recommendations for alerting and monitoring are listed. All metrics
referenced below are using the --new-metrics option, eventually the Collector
will transition to these metrics as default and disable the old ones.
Use rate of otelcol_processor_dropped_spans > 0 and
otelcol_processor_dropped_metric_points > 0 to detect data loss, depending on
the requirements set up a minimal time window before alerting, avoiding
notifications for small losses that are not considered outages or within the
desired reliability level.
This depends on the CPU metrics available on the deployment, eg.:
kube_pod_container_resource_limits_cpu_cores for Kubernetes. Let's call it
available_cores below. The idea here is to have an upper bound of the number
of available cores, and the maximum expected ingestion rate considered safe,
let's call it safe_rate, per core. This should trigger increase of resources/
instances (or raise an alert as appropriate) whenever
(actual_rate/available_cores) < safe_rate.
The safe_rate depends on the specific configuration being used.
// TODO: Provide reference safe_rate for a few selected configurations.
The queued_retry processor is recommended as the retry mechanism for the
Collector and as such should be used in any production deployment.
The queued_retry processor provides the
otelcol_processor_queued_retry_queue_length metric, besides others.
When this metric is growing constantly it is an indication that the Collector
is not able to send data as fast as it is receiving.
This will precede data loss and also can indicate a Collector low on resources.
Sustained rates of otelcol_receiver_refused_spans and
otelcol_receiver_refused_metric_points indicate too many errors returned to
clients. Depending on the deployment and the client’s resilience this may
indicate data loss at the clients.
Sustained rates of otelcol_exporter_send_failed_spans and
otelcol_exporter_send_failed_metric_points indicate that the Collector is not
able to export data as expected.
It doesn't imply data loss per se since there could be retries but a high rate
of failures could indicate issues with the network or backend receiving the
data.
The otelcol_receiver_accepted_spans and
otelcol_receiver_accepted_metric_points metrics provide information about
the data ingested by the Collector.
The otecol_exporter_sent_spans and
otelcol_exporter_sent_metric_pointsmetrics provide information about
the data exported by the Collector.