This is an automated email from the ASF dual-hosted git repository.
yjhjstz pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/cloudberry.git
commit 16ff096c3eea3911fe2dcbfd0b0c8d3eb8a7224b
Author: Huansong Fu <fuhuans...@gmail.com>
AuthorDate: Tue Jun 20 12:42:23 2023 -0700
Fix gp_stat_bgwriter which shows incomplete/incorrect results
The gp_stat_bgwriter view which is supposed to show cluster-wide view
of pg_stat_bgwriter but only shows part of the segments, e.g.:
```sql
postgres=# select * from gp_stat_bgwriter;
segment_id | checkpoints_timed | checkpoints_req | checkpoint_write_time |
checkpoint_sync_time | buffers_checkpoint | buffers_clean | maxwritten_clean |
buffers_backend | buffers_backend_fsync | buffers_alloc | stats_reset
------------+-------------------+-----------------+-----------------------+----------------------+--------------------+---------------+------------------+-----------------+-----------------------+---------------+-------------------------------
0 | 1097 | 0 | 37946 |
15 | 378 | 0 | 0 |
5133 | 0 | 387 | 2023-04-28
12:46:24.938582-07
-1 | 1097 | 0 | 37946 |
15 | 378 | 0 | 0 |
5133 | 0 | 387 | 2023-04-28
12:46:24.938582-07
(2 rows)
```
And even the result of the only primary segment is not correct - it is
just repeating the result from the coordinator.
The root cause is that because pg_stat_bgwriter is selecting from a bunch
of STABLE functions, they are being evaluated and simplied in planning
phase.
As a result, the coordinator's results are incorrectly used for the primary.
In addition, the subquery SELECT * FROM pg_stat_bgwriter is given a locus
CdbLocusType_General which results in direct dispatch and only one primary
is being dispatched to.
The initial fix that I thought was to not simplify stable functions at
all, but that would be an overkill. And it's not easy to do something like
in #14499 where we skip the simpifying just for the functions that
pg_stat_bgwriter needs - there are too many functions and no clear pattern
to
distinguish them.
So just do a simpler fix which is to have a volatile function to select the
results, and let pg_stat_bgwriter select from that function.
Also modified the regress test to be able to catch such issue in future.
Found a new issue with pg_replication_slot which is unrelated to the current
one. Added a FIXME for it.
Fix #15455
---
src/backend/catalog/system_views.sql | 46 ++++++++++++++++++++++++++----------
1 file changed, 34 insertions(+), 12 deletions(-)
diff --git a/src/backend/catalog/system_views.sql
b/src/backend/catalog/system_views.sql
index 561f89de8f..7ec1293af3 100644
--- a/src/backend/catalog/system_views.sql
+++ b/src/backend/catalog/system_views.sql
@@ -1545,19 +1545,41 @@ CREATE VIEW pg_stat_archiver AS
s.stats_reset
FROM pg_stat_get_archiver() s;
-CREATE VIEW pg_stat_bgwriter AS
+-- Internal function for pg_stat_bgwriter. It needs to be VOLATILE in order
+-- for pg_stat_bgwriter to work correctly with gp_dist_random.
+CREATE OR REPLACE FUNCTION pg_stat_bgwriter_func()
+RETURNS TABLE (
+ checkpoints_timed BIGINT,
+ checkpoints_req BIGINT,
+ checkpoint_write_time FLOAT,
+ checkpoint_sync_time FLOAT,
+ buffers_checkpoint BIGINT,
+ buffers_clean BIGINT,
+ maxwritten_clean BIGINT,
+ buffers_backend BIGINT,
+ buffers_backend_fsync BIGINT,
+ buffers_alloc BIGINT,
+ stats_reset TIMESTAMPTZ
+)
+AS
+$$
SELECT
- pg_stat_get_bgwriter_timed_checkpoints() AS checkpoints_timed,
- pg_stat_get_bgwriter_requested_checkpoints() AS checkpoints_req,
- pg_stat_get_checkpoint_write_time() AS checkpoint_write_time,
- pg_stat_get_checkpoint_sync_time() AS checkpoint_sync_time,
- pg_stat_get_bgwriter_buf_written_checkpoints() AS buffers_checkpoint,
- pg_stat_get_bgwriter_buf_written_clean() AS buffers_clean,
- pg_stat_get_bgwriter_maxwritten_clean() AS maxwritten_clean,
- pg_stat_get_buf_written_backend() AS buffers_backend,
- pg_stat_get_buf_fsync_backend() AS buffers_backend_fsync,
- pg_stat_get_buf_alloc() AS buffers_alloc,
- pg_stat_get_bgwriter_stat_reset_time() AS stats_reset;
+ pg_stat_get_bgwriter_timed_checkpoints(),
+ pg_stat_get_bgwriter_requested_checkpoints(),
+ pg_stat_get_checkpoint_write_time(),
+ pg_stat_get_checkpoint_sync_time(),
+ pg_stat_get_bgwriter_buf_written_checkpoints(),
+ pg_stat_get_bgwriter_buf_written_clean(),
+ pg_stat_get_bgwriter_maxwritten_clean(),
+ pg_stat_get_buf_written_backend(),
+ pg_stat_get_buf_fsync_backend(),
+ pg_stat_get_buf_alloc(),
+ pg_stat_get_bgwriter_stat_reset_time();
+$$
+LANGUAGE SQL;
+
+CREATE VIEW pg_stat_bgwriter AS
+ SELECT * FROM pg_stat_bgwriter_func();
CREATE VIEW pg_stat_wal AS
SELECT
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@cloudberry.apache.org
For additional commands, e-mail: commits-h...@cloudberry.apache.org
