Add new compute metrics to sql exporter (#9190)

These are the perf counters added in commit 263dfba6ee.

Note: This relies on 'neon' extension version 1.5. The default was
bumped to 1.5 in commit d696c41807.

---------

Co-authored-by: Matthias van de Meent <matthias@neon.tech>
This commit is contained in:
Heikki Linnakangas
2024-10-01 17:38:19 +03:00
committed by GitHub
parent 1b8b50755c
commit 0d500bbd5b
2 changed files with 86 additions and 1 deletions

View File

@@ -94,6 +94,68 @@ metrics:
query: |
select sum(pg_database_size(datname)) as total from pg_database;
- metric_name: getpage_wait_seconds_count
type: counter
help: 'Number of getpage requests'
values: [getpage_wait_seconds_count]
query_ref: neon_perf_counters
- metric_name: getpage_wait_seconds_sum
type: counter
help: 'Time spent in getpage requests'
values: [getpage_wait_seconds_sum]
query_ref: neon_perf_counters
- metric_name: getpage_prefetch_requests_total
type: counter
help: 'Number of getpage issued for prefetching'
values: [getpage_prefetch_requests_total]
query_ref: neon_perf_counters
- metric_name: getpage_sync_requests_total
type: counter
help: 'Number of synchronous getpage issued'
values: [getpage_sync_requests_total]
query_ref: neon_perf_counters
- metric_name: getpage_prefetch_misses_total
type: counter
help: 'Total number of readahead misses; consisting of either prefetches that don't satisfy the LSN bounds once the prefetch got read by the backend, or cases where somehow no readahead was issued for the read'
values: [getpage_prefetch_misses_total]
query_ref: neon_perf_counters
- metric_name: getpage_prefetch_discards_total
type: counter
help: 'Number of prefetch responses issued but not used'
values: [getpage_prefetch_discards_total]
query_ref: neon_perf_counters
- metric_name: pageserver_requests_sent_total
type: counter
help: 'Number of all requests sent to the pageserver (not just GetPage requests)'
values: [pageserver_requests_sent_total]
query_ref: neon_perf_counters
- metric_name: pageserver_disconnects_total
type: counter
help: 'Number of times that the connection to the pageserver was lost'
values: [pageserver_disconnects_total]
query_ref: neon_perf_counters
- metric_name: pageserver_send_flushes_total
type: counter
help: 'Number of flushes to the pageserver connection'
values: [pageserver_send_flushes_total]
query_ref: neon_perf_counters
- metric_name: getpage_wait_seconds_buckets
type: counter
help: 'Histogram buckets of getpage request latency'
key_labels:
- bucket_le
values: [value]
query_ref: getpage_wait_seconds_buckets
# DEPRECATED
- metric_name: lfc_approximate_working_set_size
type: gauge
@@ -244,3 +306,26 @@ metrics:
SELECT slot_name,
CASE WHEN wal_status = 'lost' THEN 1 ELSE 0 END AS wal_is_lost
FROM pg_replication_slots;
queries:
- query_name: neon_perf_counters
query: |
WITH c AS (
SELECT pg_catalog.jsonb_object_agg(metric, value) jb FROM neon.neon_perf_counters
)
SELECT d.*
FROM pg_catalog.jsonb_to_record((select jb from c)) as d(
getpage_wait_seconds_count numeric,
getpage_wait_seconds_sum numeric,
getpage_prefetch_requests_total numeric,
getpage_sync_requests_total numeric,
getpage_prefetch_misses_total numeric,
getpage_prefetch_discards_total numeric,
pageserver_requests_sent_total numeric,
pageserver_disconnects_total numeric,
pageserver_send_flushes_total numeric
);
- query_name: getpage_wait_seconds_buckets
query: |
SELECT bucket_le, value FROM neon.neon_perf_counters WHERE metric = 'getpage_wait_seconds_bucket';

View File

@@ -137,7 +137,7 @@ neon_perf_counters_to_metrics(neon_per_backend_counters *counters)
metrics[i].is_bucket = false;
metrics[i].value = (double) counters->pageserver_requests_sent_total;
i++;
metrics[i].name = "pageserver_requests_disconnects_total";
metrics[i].name = "pageserver_disconnects_total";
metrics[i].is_bucket = false;
metrics[i].value = (double) counters->pageserver_disconnects_total;
i++;