Re: Add sub-transaction overflow status in pg_stat_activity

On Tue, Dec 14, 2021 at 6:23 PM Ashutosh Sharma <ashu(dot)coek88(at)gmail(dot)com>
wrote:

In the latest patch I have fixed comments given here except a few.

I have looked into the v2 patch and here are my comments:

+ PG_RETURN_INT32(local_beentry->subxact_overflowed);
+}

Should this be PG_RETURN_BOOL instead of PG_RETURN_INT32??

With the new patch this is not relevant because we are returning a tuple.

+{ oid => '6107', descr => 'statistics: cached subtransaction count of
backend',
+ proname => 'pg_stat_get_backend_subxact_count', provolatile => 's',
proparallel => 'r',
+ prorettype => 'int4', proargtypes => 'int4',
+ prosrc => 'pg_stat_get_backend_subxact_count' },
+{ oid => '6108', descr => 'statistics: subtransaction cache of backend
overflowed',
+ proname => 'pg_stat_get_backend_subxact_overflow', provolatile => 's',
proparallel => 'r',
+ prorettype => 'bool', proargtypes => 'int4',
+ prosrc => 'pg_stat_get_backend_subxact_overflow' },

The description says that the two new functions show the statistics for
"cached subtransaction count of backend" and "subtransaction cache of
backend overflowed". But, when these functions are called it shows these
stats for the non-backend processes like checkpointer, walwriter etc as
well. Should that happen?

I am following similar description as pg_stat_get_backend_pid,
pg_stat_get_backend_idset and other relevant functioins.

typedef struct LocalPgBackendStatus
* not.
*/
TransactionId backend_xmin;
+
+ /*
+ * Number of cached subtransactions in the current session.
+ */
+ int subxact_count;
+
+ /*
+ * The number of subtransactions in the current session exceeded the
cached
+ * subtransaction limit.
+ */
+ bool subxact_overflowed;

All the variables inside this LocalPgBackendStatus structure are prefixed
with "backend" word. Can we do the same for the newly added variables as
well?

Done

+ * Get the xid and xmin, nsubxid and overflow status of the backend.
The

Should this be put as - "xid, xmin, nsubxid and overflow" instead of "xid
and xmin, nsubxid and overflow"?

I missed to fix this one in the last patch so updated again.

--
Regards,
Dilip Kumar
EnterpriseDB: http://www.enterprisedb.com