Add metric for monitoring for blocked queries (#207)

exporter/postgres/queries_pg10.yml

@@ -11,6 +11,7 @@ ccp_connection_stats:
         , idle_in_txn
         , (select coalesce(extract(epoch from (max(now() - state_change))),0) from pg_catalog.pg_stat_activity where state = 'idle in transaction') as max_idle_in_txn_time
         , (select coalesce(extract(epoch from (max(now() - query_start))),0) from pg_catalog.pg_stat_activity where backend_type = 'client backend' and state <> 'idle' ) as max_query_time
+        , (select coalesce(extract(epoch from (max(now() - query_start))),0) from pg_catalog.pg_stat_activity where backend_type = 'client backend' and wait_event_type = 'Lock' ) as max_blocked_query_time
         , max_connections
         from (
                 select count(*) as total
@@ -36,6 +37,9 @@ ccp_connection_stats:
     - max_query_time:
         usage: "GAUGE"
         description: "Length of time in seconds of the longest running query"
+    - max_blocked_query_time:
+        usage: "GAUGE"
+        description: "Length of time in seconds of the longest running query that has been blocked by a heavyweight lock"
     - max_connections:
         usage: "GAUGE"
         description: "Value of max_connections for the monitored database"

exporter/postgres/queries_pg11.yml

@@ -11,6 +11,7 @@ ccp_connection_stats:
         , idle_in_txn
         , (select coalesce(extract(epoch from (max(now() - state_change))),0) from pg_catalog.pg_stat_activity where state = 'idle in transaction') as max_idle_in_txn_time
         , (select coalesce(extract(epoch from (max(now() - query_start))),0) from pg_catalog.pg_stat_activity where backend_type = 'client backend' and state <> 'idle' ) as max_query_time
+        , (select coalesce(extract(epoch from (max(now() - query_start))),0) from pg_catalog.pg_stat_activity where backend_type = 'client backend' and wait_event_type = 'Lock' ) as max_blocked_query_time
         , max_connections
         from (
                 select count(*) as total
@@ -36,6 +37,9 @@ ccp_connection_stats:
     - max_query_time:
         usage: "GAUGE"
         description: "Length of time in seconds of the longest running query"
+    - max_blocked_query_time:
+        usage: "GAUGE"
+        description: "Length of time in seconds of the longest running query that has been blocked by a heavyweight lock"
     - max_connections:
         usage: "GAUGE"
         description: "Value of max_connections for the monitored database"

exporter/postgres/queries_pg12.yml

@@ -11,6 +11,7 @@ ccp_connection_stats:
         , idle_in_txn
         , (select coalesce(extract(epoch from (max(now() - state_change))),0) from pg_catalog.pg_stat_activity where state = 'idle in transaction') as max_idle_in_txn_time
         , (select coalesce(extract(epoch from (max(now() - query_start))),0) from pg_catalog.pg_stat_activity where backend_type = 'client backend' and state <> 'idle' ) as max_query_time
+        , (select coalesce(extract(epoch from (max(now() - query_start))),0) from pg_catalog.pg_stat_activity where backend_type = 'client backend' and wait_event_type = 'Lock' ) as max_blocked_query_time
         , max_connections
         from (
                 select count(*) as total
@@ -36,6 +37,9 @@ ccp_connection_stats:
     - max_query_time:
         usage: "GAUGE"
         description: "Length of time in seconds of the longest running query"
+    - max_blocked_query_time:
+        usage: "GAUGE"
+        description: "Length of time in seconds of the longest running query that has been blocked by a heavyweight lock"
     - max_connections:
         usage: "GAUGE"
         description: "Value of max_connections for the monitored database"

exporter/postgres/queries_pg13.yml

@@ -11,6 +11,7 @@ ccp_connection_stats:
         , idle_in_txn
         , (select coalesce(extract(epoch from (max(now() - state_change))),0) from pg_catalog.pg_stat_activity where state = 'idle in transaction') as max_idle_in_txn_time
         , (select coalesce(extract(epoch from (max(now() - query_start))),0) from pg_catalog.pg_stat_activity where backend_type = 'client backend' and state <> 'idle' ) as max_query_time
+        , (select coalesce(extract(epoch from (max(now() - query_start))),0) from pg_catalog.pg_stat_activity where backend_type = 'client backend' and wait_event_type = 'Lock' ) as max_blocked_query_time
         , max_connections
         from (
                 select count(*) as total
@@ -36,6 +37,9 @@ ccp_connection_stats:
     - max_query_time:
         usage: "GAUGE"
         description: "Length of time in seconds of the longest running query"
+    - max_blocked_query_time:
+        usage: "GAUGE"
+        description: "Length of time in seconds of the longest running query that has been blocked by a heavyweight lock"
     - max_connections:
         usage: "GAUGE"
         description: "Value of max_connections for the monitored database"

exporter/postgres/queries_pg95.yml

@@ -11,6 +11,7 @@ ccp_connection_stats:
         , idle_in_txn
         , (select coalesce(extract(epoch from (max(now() - state_change))),0) from monitor.pg_stat_activity() where state = 'idle in transaction') as max_idle_in_txn_time
         , (select coalesce(extract(epoch from (max(now() - query_start))),0) from monitor.pg_stat_activity() where state <> 'idle') as max_query_time
+        , (select coalesce(extract(epoch from (max(now() - query_start))),0) from monitor.pg_stat_activity() where waiting = 't' ) as max_blocked_query_time
         , max_connections
         from (
                 select count(*) as total
@@ -36,6 +37,9 @@ ccp_connection_stats:
     - max_query_time:
         usage: "GAUGE"
         description: "Length of time in seconds of the longest running query"
+    - max_blocked_query_time:
+        usage: "GAUGE"
+        description: "Length of time in seconds of the longest running query that has been blocked by a heavyweight lock"
     - max_connections:
         usage: "GAUGE"
         description: "Value of max_connections for the monitored database"

exporter/postgres/queries_pg96.yml

@@ -12,6 +12,7 @@ ccp_connection_stats:
         , idle_in_txn
         , (select coalesce(extract(epoch from (max(now() - state_change))),0) from monitor.pg_stat_activity() where state = 'idle in transaction') as max_idle_in_txn_time
         , (select coalesce(extract(epoch from (max(now() - query_start))),0) from monitor.pg_stat_activity() where state <> 'idle') as max_query_time
+        , (select coalesce(extract(epoch from (max(now() - query_start))),0) from pg_catalog.pg_stat_activity where backend_type = 'client backend' and wait_event_type = 'Lock' ) as max_blocked_query_time
         , max_connections
         from (
                 select count(*) as total
@@ -37,6 +38,9 @@ ccp_connection_stats:
     - max_query_time:
         usage: "GAUGE"
         description: "Length of time in seconds of the longest running query"
+    - max_blocked_query_time:
+        usage: "GAUGE"
+        description: "Length of time in seconds of the longest running query that has been blocked by a heavyweight lock"
     - max_connections:
         usage: "GAUGE"
         description: "Value of max_connections for the monitored database"

hugo/content/changelog/_index.md

@@ -9,6 +9,7 @@ weight: 5
 
   * Added support for PostgreSQL 13 
   * Added queries and dashboards for pgnodemx/container support
+  * Added metrics for monitoring longest blocked query time
 
 ### Bug Fixes
 

hugo/content/exporter/_index.md

@@ -507,6 +507,8 @@ The following metrics either require special considerations when monitoring spec
 
  * *ccp_connection_stats_idle_in_txn* - Count of idle in transaction connections
 
+ * *ccp_connection_stats_max_blocked_query_time* - Runtime of longest running query that has been blocked by a heavyweight lock
+
  * *ccp_connection_stats_max_connections* - Current value of max_connections for reference
 
  * *ccp_connection_stats_max_idle_in_txn_time* - Runtime of longest idle in transaction (IIT) session.