I see this on 5.0.35-1 on Ubuntu too

Same problem here. Our logs outputs are identical to yours.

It seems like the PG.USER macro is in the database field....

Zabbix-agent2 1:6.0.18-1+debian10 amd64

Ok, I tried to fix the authentication part (dropping the role zbx_monitor and creating a zabbix one) but, as it could be expected (reading accurately the logs above) now the plugin query a evidently non-existent db "user=zabbix".

Please see the attached postgres and zabbix-agent2 (level 3) debug log ZBX-22868.log.

Maybe the query's parameters are not passed correctly?

PS: <zabbix_server_ip_address>, <pgsql_password> replace the actual values for privacyZBX-22868.log

I think that the only relevant work done between 6.0.17 and 6.0.18 is https://support.zabbix.com/browse/ZBXNEXT-8330

But I don't know how to relate that to this issue...

Same issue on 6.4.3 - server and agent was installed with 6.4.2 where it worked fine, then after todays upgrade to 6.4.3 it breaks.

Seemingly something to do with  ZBXNEXT-8330 where defaults are now applied instead of macros, how the database ends up being "user=zbx_monitor" I am not sure.

Connection failed: failed to connect to `host=localhost user=zabbix database=user=zbx_monitor

Full zabbix-agent2 log line:

2023/06/01 12:00:53.816001 [PostgreSQL] Connection failed: failed to connect to `host=localhost user=zabbix database=user=nagios`: failed SASL auth (FATAL: password authentication failed for user "zabbix" (SQLSTATE 28P01)).

We use a nagios user (as set in the template macro) to connect to the PostgreSQL database.

Updating the postgres plugin but holding the zabbix agent on 6.4.2 produces the same results.

Temporary solution is to modify plugin config:

Plugins.PostgreSQL.Default.Uri=tcp://localhost:5432
Plugins.PostgreSQL.Default.User=XXXXX
Plugins.PostgreSQL.Default.Password=YYYYY
Plugins.PostgreSQL.Default.Database=postgres

and set empty macros:

{$PG.URI}
{$PG.USER}
{$PG.PASSWORD}

Server on debian 11, zabbix server 6.4.3

Client on debian 11,ztabbix-agent2 and zabbix-agent2-plugin-postgresql.

Works with macro in server due to need of monitoring multiple instance on same machine.

Previous release: 6.4.2-1+debian11

Update to: 6.4.3-1+debian11

Break the monitoring with log:

1. request received

created direct exporter task for plugin 'PostgreSQL' itemid:0 key 'pgsql.dbstat.sum["tcp://127.0.0.1:5432","USERNAME","PASSWORD"]'

created direct exporter task for plugin 'PostgreSQL' itemid:0 key 'pgsql.dbstat.sum["tcp://127.0.0.1:5432","USERNAME","PASSWORD"]'

1. error on

failed to execute direct exporter task for key 'pgsql.dbstat.sum["tcp://127.0.0.1:5432","USERNAME","PASSWORD"]' error: '
Connection failed: failed to connect to `host=127.0.0.1 user=zabbix database=user=USERNAME`

Looks like plugin find no user and use user=zabbix, and use for db user=USERNAME

Back to version 6.4.2-1+debian11. Everything works fine again.

Experiencing the same.

Looking at PostgreSQL logs:

2023-06-03 22:48:44.669 UTC [52466] zabbix@user=zbx_monitor FATAL:  password authentication failed for user "zabbix"
2023-06-03 22:48:44.669 UTC [52466] zabbix@user=zbx_monitor DETAIL:  Connection matched pg_hba.conf line 101: "host    all             all             127.0.0.1/32            scram-sha-256" 

we can say that plugin suddenly uses user "zabbix" to connect to "user=zbx_monitor" database.

I really wonder how this one made it to the final 6.0.18 release as the problem was already reported with 6.0.18rc1?

At least the testing workflow may be improved for future releases?

Correct me if I'm wrong but the only "real" solution for now is to downgrade to 6.0.17 and wait for a fixed (and hopefully better tested) 6.0.19.

Hello @JohnWayne , please note that i reported the bug on 6.0.18rc1 but i have discovered it when upgrading to 6.0.18 via deb packaging.

So the bug report has been made after the release of 6.0.18. I cannot recall now why I wrote 6.0.18rc1, probably that was the closest option available in the combo box .

@Tomasz Berner proposed a temporary workaround that i did not try yet, though.

As a temporary workaround adding Plugins.PostgreSQL.Default.Database=postgres to agent configuration file should be enough.

We were working fine on 6.0.17, the upgrade broke posgres monitoring. We just installed the latest patch and still get errors.

failed: failed to connect to host=localhost user=zabbix database=user=zbx_monitor: failed SASL auth (FATAL: password authentication failed for user "zabbix" (SQLSTATE 28P01)).' to 'X.X.X.X'

On RHEL latest work zabbix-agent2 and PostgreSQL plugin 6.4.0 version.

Fixed in:

• Zabbix main repository
  • 5.0.36rc1: 2c3d18f8e15
  • 6.0.19rc1: bb542491adc
  • 6.4.4rc1: 65e30906f02
  • 7.0.0alpha2 (master): 92c199a76a6
• Plugin Support
  • 6.0.19rc1: 45b1778277a
  • 6.4.4rc1: eaed6a1f2e7
  • 7.0.0alpha2 (master): 1f7180b85c3
• MongoDB
  • 6.0.19rc1: 1286080d785
  • 6.4.4rc1: b0e1aba8c5a
  • 7.0.0alpha2 (master): a200df97983
• PostgreSQL
  • 6.0.19rc1: 4efda8efd49
  • 6.4.4rc1: 2605da73f30
  • 7.0.0alpha2 (master): 2c54a34f321