Bump pgmonitor to v5.2.1 #4186
Bump pgmonitor to v5.2.1 #4186
Conversation
| SELECT c.buffers_checkpoint AS buffers_written | ||
| FROM pg_catalog.pg_stat_bgwriter c; | ||
| metrics: | ||
| - metric_name: ccp_stat_bgwriter_buffers_checkpoint |
There was a problem hiding this comment.
When did the ccp_stat_bgwriter_* metrics go away?
There was a problem hiding this comment.
Not all ccp_stat_bgwriter_* metrics went away, but in pgMonitor v5.1.0 I think, ones that were "checkpoint" related moved to ccp_stat_checkpointer and then in pgmonitor-extension, the function checks if the PG version is gte 17, and if so, it uses pg_stat_checkpointer and if it's less than 17 it uses pg_stat_bgwriter.
| DROP FUNCTION IF EXISTS get_replication_lag(); | ||
| --- get_replication_lag is used by the OTel collector. | ||
| --- get_replication_lag is created as function, so that we can query without warning on a replica. | ||
| CREATE OR REPLACE FUNCTION get_replication_lag() RETURNS TABLE(bytes NUMERIC) AS $$ | ||
| CREATE FUNCTION get_replication_lag() RETURNS TABLE(replica text, bytes NUMERIC) AS $$ |
There was a problem hiding this comment.
🤔 Is there a reason to prefer "drop/create" vs. "create or replace"?
There was a problem hiding this comment.
When I changed some things inside the function I started getting errors with CREATE OR REPLACE and it said I needed to DROP the function instead.
There was a problem hiding this comment.
Interesting, I wonder what I'm missing -- I wonder if "CREATE OR REPLACE" can't actually change some aspects....
Ah well, not a problem here, just something I was curious about.
There was a problem hiding this comment.
Yeah it definitely doesn't like to replace the function if the signature changes but I think it also complained when other non-signature related things changed...
There was a problem hiding this comment.
LGTM, AFAICT -- I used a compare between 5.1.1 and 5.2.1 in pgmonitor to check that we have all the changes and one thing I noticed is that we don't have a ccp_autovacuum_workers metric. (I only noticed that because they moved it btw 5.1.1 and 5.2.1.) Is this not a metric we use?
I don't see any references to |
Combine ccp_archive_command_status queries into one query. Add semicolons to the end of all queries. Make ccp_replication_lag_size return the replica name for grafana dashboard legend. DROP functions rather than CREATE OR REPLACE to avoid errors due to changes in functions.
dsessler7
force-pushed
the
bump-pgmonitor-to-v5.2.1
branch
2 times, most recently
from
77bdfa5 to
da63454
Compare
Checklist:
Type of Changes:
What is the current behavior (link to any open issues here)?
Our OTel metrics currently use a mixture of metrics from pgMonitor v5.1.1 and earlier.
What is the new behavior (if this is a feature change)?
This PR brings most of the metrics up to their equivalent implementations in pgMonitor v5.2.1.
Other Information: