Monitoring

pgec is instrumented using telemetry with a Prometheus http adapter listening on port 9100.

Grafana dashboards: http://localhost:3000/.

The following sections describe the metrics that are exposed by pgec.

BEAM

Metrics for the BEAM itself:

• beam_memory, is a gauge recording memory usage in bytes classified by type..
• beam_ports, is a gauge recording file descriptor usage.
• beam_processes, is a gauge recording the number and various types of processes.

Storage Cache

The storage cache is a memory KV cache (implemented using ETS), automatically expiring unused values.

• pgec_storage_cache, labelled with the action (hit, miss, expired, delete, update, insert)

Storage Backend

Storage requests that are processed by the storage backend (leveled). Reads that are as a result of a cache miss, and all writes will be processed by the storage backend:

• pgec_storage_delete_count, requests to delete a key from the store.
• pgec_storage_get_count, requests to read a value represented by a key from the store.
• pgec_storage_put_count, requests to associate a key with a value.

redis compatible API endpoint

Metrics for the redis compatible protocol subsystem:

• resp_listener_accept, resp_listener_bind, resp_listener_connect resp_listener_listen, resp_listener_open and resp_listener_select, are all listener socket state metrics.
• resp_connection_connect, resp_connection_decode, resp_connection_encode, resp_connection_open, resp_connection_recv, resp_connection_send, are connection metrics including protocol decoding/encoding of messages, and the sending or receiving of protocol data.

memcached compatible API endpoint

Metrics for the memcached protocol subsystem:

• mcd_tcp_connection_recv_bytes, in the memcached subsystem, the number of bytes that have been received.
• mcd_tcp_connection_send_bytes, in the memcached subsystem, the number of bytes that have been sent.

REST endpoint

Metrics for the REST endpoint:

• cowboy_request_count, the number of HTTP requests made.
• cowboy_request_duration_ms, cumulative milliseconds taken processing all requests.
• cowboy_request_resp_body_length, the cumulative length of all response bodies.

PostgreSQL replication

Metrics for PostgreSQL replication:

• pgmp_mm_execute_rows, includes the number of rows processed during initial data collection, prior to streaming replication starting.
• pgmp_mm_rep_keepalive_count, a counter of the keep alive requests that have been processed.
• pgmp_mm_rep_wal_applied, the index of the WAL that has been applied.
• pgmp_mm_rep_wal_clock, the latest clock value received from the WAL.
• pgmp_mm_rep_wal_flushed, the index of the WAL that has been flushed.
• pgmp_mm_rep_wal_received, the latest index of the WAL that been received.
• pgmp_mm_rep_begin_transaction_count, a counter of the number of transactions that have begun.
• pgmp_mm_rep_commit_count, a counter of the number of transactions that have been committed.
• pgmp_mm_rep_insert_count, a counter of the number of rows that have been inserted during streaming replication.
• pgmp_mm_rep_update_count, a counter of the number of rows that have been updated during streaming replication.
• pgmp_mm_rep_delete_count, a counter of the number of rows that have been deleted during streaming replication.
• pgmp_mm_rep_truncate_count, a counter of the number table truncates that have been processed.