Merge pull request #15 from ongres/pg16

add queries to analyze IO statistics for PG16

sql/IO/IO.md

@@ -0,0 +1,11 @@
+# IO analysis
+
+Taking advantage from the new view [pg_stat_io](https://www.postgresql.org/docs/devel/monitoring-stats.html#MONITORING-PG-STAT-IO-VIEW) released in PG16 , it is possible to see statitics about backend type and IO operations
+
+The queries for this porpose can be used to check the information about:
+
+
+* Analyze the impact of vacuum in IO activity (high is bad,maybe >10% ?)
+* Analyze the impact of client backend in fsync activity (high is bad,maybe >10% ?)
+
+TODO: analyze , top IO consumer, top evictions

sql/IO/client_backend_fsync_impact.sql

@@ -0,0 +1,13 @@
+WITH cte_fsync_client_backend_io AS
+  (SELECT sum(fsyncs) fsync_client_backend_io
+   FROM pg_stat_io
+   WHERE backend_type = 'client backend'
+ ),
+     cte_fsync_total_io AS
+  (SELECT sum(fsyncs) fsync_total_io
+   FROM pg_stat_io)
+SELECT round((
+          (SELECT fsync_client_backend_io
+           FROM cte_fsync_client_backend_io) *100)/
+  (SELECT fsync_total_io
+   FROM cte_fsync_total_io),2) as io_backend_fsync_activity_pct;

sql/IO/vacuum_io_impact.sql

@@ -0,0 +1,16 @@
+WITH cte_vacuum_io AS
+  (SELECT sum(READS)+sum(writes)+sum(extends) vacuum_io
+   FROM pg_stat_io
+   WHERE backend_type = 'autovacuum worker'
+     OR (context = 'vacuum'
+         AND (READS <> 0
+              OR writes <> 0
+              OR extends <> 0)) ),
+     cte_total_io AS
+  (SELECT sum(READS)+sum(writes)+sum(extends) total_io
+   FROM pg_stat_io)
+SELECT round((
+          (SELECT vacuum_io
+           FROM cte_vacuum_io) *100)/
+  (SELECT total_io
+   FROM cte_total_io),2) as io_vacuum_activity_pct;