Success callback about collector

[lfittl]

@jglapa Interesting idea!

I wonder if you could check the modified time of the state file (typically at /var/lib/pganalyze-collector/state) and alert if its older than e.g. 15 minutes?

[jglapa]

From what I see the file /var/lib/pganalyze-collector/state is getting modified every run although the data is not submitted to pganalyze.

I'm currently troubleshooting the following problem:

• I have a postgresql 9.6 installation on AWS RDS
• I've configured everything as described in the manual
• I'm using the newest collector
• but when I run the collector with --dry-run I'm getting the following:

2017/04/21 15:07:25 V [xxx] Detected PostgreSQL Version 90601 (PostgreSQL 9.6.1 on x86_64-pc-linux-gnu, compiled by gcc (GCC) 4.8.2 20140120 (Red Hat 4.8.2-16), 64-bit)
2017/04/21 15:07:25 V [xxx] Found pganalyze.get_stat_statements() stats helper
2017/04/21 15:07:25 E [xxx] Error collecting pg_stat_statements
2017/04/21 15:07:25 E [xxx] Could not process database: pq: pg_stat_statements must be loaded via shared_preload_libraries

This must be an RDS issue since I've configured the shared_preload_libraries=pg_state_statements correctly in parameter group but psql shows rdsutils... ?,

([email protected]:5432) [test]  [] > \dx
                                     List of installed extensions
        Name        | Version |   Schema   |                        Description
--------------------+---------+------------+-----------------------------------------------------------
 pg_stat_statements | 1.4     | public     | track execution statistics of all SQL statements executed
 plpgsql            | 1.0     | pg_catalog | PL/pgSQL procedural language
(2 rows)


([email protected]:5432) [test]  [] > show shared_preload_libraries;
 shared_preload_libraries
--------------------------
 rdsutils
(1 row)

([email protected]:5432) [test]  [] > select * from pg_stat_statements ;
ERROR:  pg_stat_statements must be loaded via shared_preload_libraries
Time: 165.208 ms

Anyhow with this issue in place the /var/lib/pganalyze-collector/state mtime is still getting changed although an error is raised.

[lfittl]

@jglapa Fair point - for the specific issue you are seeing, first check that you have restarted your RDS server (otherwise changes to shared_preload_libraries don't take effect). We've also sometimes seen problems where RDS has an internal problem, and typically updating to the next minor PostgreSQL release fixed that.

On the success callback - I'm currently thinking we just make it --success-callback=/path/to/shell/script.sh, which gets triggered once for each server, with the first and only argument being the name of the server (as per the collector config file). That sound reasonable?

[jglapa]

--success-callback=/path/to/shell/script.sh

And how would we configure that?
This would require changing collector startup scripts (upstart right now)?

I was more thinking about a new property in the pganalyze-collector.conf file.

Passing the db config name when calling the script sound like a very good idea!

In my case I could either have 2 separate shell script( I have 2 db's) and simply ignore the passed parameter or have one script and make the distinction based on the parameter passed.