apm-aggregation

APM metrics aggregation library that implements an LSM (Log Structured Merge tree)-based metrics aggregator.

Files are subject to Elastic License v2. See LICENSE.txt for more.

Instrumentation

apm-aggregation uses OTEL to instrument itself. Instrumentation produces a set of metrics to help monitor the status of aggregations. This section describes the metrics produced by apm-aggregation in detail.

Instrumentation Areas

apm-aggregation aggregates metrics using LSM based key-value store pebble. The intrumentation covers two broad areas:

1. The core aggregation logic, including ingestion and harvest.
2. Performance of pebble database.

Metrics

apm-aggregation records and publishes the following metrics:

events.processed.count

• Type: Float64Counter

The number of processed APM Events. It includes successfully and unsuccessfully processed events, which are reported as dimensions.

Dimensions

• combined_metrics_id
• aggregation_interval
• outcome

events.processed.bytes

• Type: Int64Counter

The number of encoded bytes processed by the aggregator. This reports the same number of bytes that is written to the underlying db.

Dimensions

• combined_metrics_id
• outcome

events.processed.latency

• Type: Float64Histogram

The processing delay for a batch of APM events accepted at a specific processing time. It is recorded after removing any expected delays due to aggregation interval or configuration.

Dimensions

• combined_metrics_id
• aggregation_interval
• outcome

events.processed.queued-latency

• Type: Float64Histogram

The delay in processing a batch based on the youngest APM event received in the batch.

Dimensions

• combined_metrics_id
• aggregation_interval
• outcome

metrics.overflowed.count

• Type: Int64Counter

Estimated number of metric aggregation keys that resulted in an overflow, per interval and aggregation type.

Dimensions

• combined_metrics_id
• aggregation_interval
• aggregation_type

pebble.flushes

• Type: Int64ObservableCounter

The number of memtable flushes to disk.

pebble.flushed-bytes

• Type: Int64ObservableCounter

The number of bytes written during a flush.

pebble.compactions

• Type: Int64ObservableCounter

The number of table compactions performed by pebble.

pebble.ingested-bytes

• Type: Int64ObservableCounter

The number of bytes ingested by pebble.

pebble.compacted-bytes-read

• Type: Int64ObservableCounter

The number of bytes read during compaction.

pebble.compacted-bytes-written

• Type: Int64ObservableCounter

The number of bytes written during compaction.

pebble.memtable.total-size

• Type: Int64ObservableGauge

The current size of memtable in bytes.

pebble.disk.usage

• Type: Int64ObservableGauge

The current total disk usage by pebble in bytes, including live and obsolete files.

pebble.read-amplification

• Type: Int64ObservableGauge

The current read amplification for the db.

pebble.num-sstables

• Type: Int64ObservableGauge

The current number of SSTables.

pebble.table-readers-mem-estimate

• Type: Int64ObservableGauge

The memory in bytes used by pebble for index and fliter blocks.

pebble.estimated-pending-compaction

• Type: Int64ObservableGauge

The current number of estimated bytes pending for compaction.

pebble.marked-for-compaction-files

• Type: Int64ObservableGauge

The current number of SSTables marked for compaction.

pebble.keys.tombstone.count

• Type: Int64ObservableGauge

The approximate count of delete keys across the storage engine.

Dimensions

This section lists the general dimensions published by some of the metric.

combined_metrics_id

This is an optional dimension. The key and the value of this dimension depends on the option WithCombinedMetricsIDToKVs passed to the aggregator. If this option is not supplied then this dimension is omitted.

aggregation_interval

Holds the value of aggregation interval for which the combined metrics is produced. For example: 1m, 10m, etc.

aggregation_type

Holds the the aggregation type for which an overflow occurred. For example: service, transaction, service_transaction, service_destination.

outcome

success

Events that have been successfully aggregated into the final combined metrics and processed as part of the harvest.

failure

Events that failed to be aggregated for an reason and were dropped at any stage in the aggregation process.