[pgpool-general: 9222] Re: pgpool-ii configurations

[Ron Johnson]

On Tue, Oct 15, 2024 at 11:45 AM Long Nguyen <nguyenxl059 at gmail.com> wrote:

> Hi,
>
> I am new to pgpool-ii and am setting it up. My env:
> - postgresql 16 on ubuntu 22.04 on both primary and standby servers
> - postgresql streaming replication is use
> - install pgpool 4.5.4 on Primary server with streaming replication mode
>
> I have followed online documentation and have some general questions.
>
> 1) Since pgpool accepts client connection and passes the same credential
> supplied by client to postgresql server, do we have the same entries for
> pgpool_hba.conf as in pg_hba.conf?
>

Typically yes,  but do not have to.  Really, it depends on your PgPool
configuration.  That's because the Postgresql instance sees the server
where the PgPool VIP is located.

For example, we have a three node PgPool cluster:
node_a: the database primary server. PG and PgPool installed.
node_b: the database streaming replica server. PG and PgPool installed.
node_c: where the VIP normally lives.  Only PgPool installed.

The node_a pg_stat_activity.client_hostname field sees "node_c", not the
application server(s) and user(s).  Thus, you don't need them in the node_a
and node_b pg_hba.conf files, but you do need them in the three
pool_hba.conf files.


> 2) The documentation recommends a separate OS user to be used to run
> pgpool. Do people do that or just use root? I created a 'pgpool' linux user
> and used it, but encountered various issues with file  access permissions.
>

I had a similar question when I first tried configuring PgPool.

The RPM packages supplied by https://www.pgpool.net/yum/rpms/4.5/ all use
the same OS user "postgres" that PostgreSQL itself uses.  APT probably does
something similar.


> 3) With the user we specify in the pcp.conf file, is it an OS user or
> postgresql server user? Or is it neither? (The doco does not make this
> clear.)
>

It's a postgresql role.


> 4) When is the user specified in the pcp.conf file referred to? Is this
> when we issue PCP commands (e.g. pcp_attach_node). How about $HOME/.pcppass
> file? If .pcppass is used then the pcp.conf file would not be referred to?
>

This is the same as Postgresql storing the role's password hash in
pg_authid and the plaintext in ~/.pgpass.

-- 
Death to <Redacted>, and butter sauce.
Don't boil me, I'm still alive.
<Redacted> crustacean!
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.pgpool.net/pipermail/pgpool-general/attachments/20241015/6cd44b42/attachment.htm>