public inbox for [email protected]
help / color / mirror / Atom feedRe: Replication using mTLS issue
2+ messages / 2 participants
[nested] [flat]
* Re: Replication using mTLS issue
@ 2024-06-21 17:21 Drew Zoellner <[email protected]>
0 siblings, 1 reply; 2+ messages in thread
From: Drew Zoellner @ 2024-06-21 17:21 UTC (permalink / raw)
To: Tom Lane <[email protected]>; +Cc: pgsql-general; [email protected]
Hi Tom, thanks for the response!
So the same user is able to connect using a non replication connection
using the same mtls certificate and pg_ident.conf map. So it seems like the
cert & map are working for this user.
hostssl all pgrepmgr_nonprod 100.0.0.0/8 cert map=pgrepmgr_nonprod_map
This above seems to be the rule that matched the non replication connection
which was successful.
I have tried relaxing the pg_hba.conf line to all like you suggested for
the username and also for IPs and other combinations, unfortunately nothing
was working.
I have been sure to use SELECT pg_reload_conf(); to update changes made to
the pg_hba.conf. I have additionally used SELECT pg_hba_file_rules(); to
verify the rules are showing up as expected from the live DB perspective.
Since non replication connections are working, and the only change to HBA
conf for the replication connection is just all -> replication , it seems
like it should be matching. Any other suggestions?
Thanks, Drew.
On Fri, Jun 21, 2024 at 11:46 AM Tom Lane <[email protected]> wrote:
> Drew Zoellner <[email protected]> writes:
> > Hi Postgres team, I’m receiving an issue matching pg_hba rules that I
> can’t
> > seem to sort out. I am trying to use mtls certificate authentication for
> > physical replication connections but keep receiving the following error…
>
> > pg_receivewal: error: FATAL: no pg_hba.conf entry for replication
> > connection from host "100.84.12.223", user "pgrepmgr_nonprod", SSL on
>
> > My pg_hba.conf file contains
> > hostssl replication pgrepmgr_nonprod 100.0.0.0/8 cert
> map=pgrepmgr_nonprod_map
>
> Hm, the match failure must be on user name. What certificate are you
> using on the client side, and what user name does pgrepmgr_nonprod_map
> map it to? Does it succeed if you weaken the hba entry to
>
> hostssl replication all 100.0.0.0/8 cert map=pgrepmgr_nonprod_map
>
> > Is cert authentication supported for replication connections?
>
> Should be. But you might find it easier to debug the auth failure
> in a non-replication context, ie add
>
> hostssl all pgrepmgr_nonprod 100.0.0.0/8 cert
> map=pgrepmgr_nonprod_map
>
> and then see if you can connect with the same credentials from psql
> or your favorite other client.
>
> BTW, don't forget you have to signal the postmaster to reload
> configuration after any change in these files.
>
> regards, tom lane
>
^ permalink raw reply [nested|flat] 2+ messages in thread
* Re: Replication using mTLS issue
@ 2024-06-21 18:24 Tom Lane <[email protected]>
parent: Drew Zoellner <[email protected]>
0 siblings, 0 replies; 2+ messages in thread
From: Tom Lane @ 2024-06-21 18:24 UTC (permalink / raw)
To: Drew Zoellner <[email protected]>; +Cc: pgsql-general; [email protected]
Drew Zoellner <[email protected]> writes:
> So the same user is able to connect using a non replication connection
> using the same mtls certificate and pg_ident.conf map. So it seems like the
> cert & map are working for this user.
Hmph. I tried to reproduce your problem, and it works for me: I can
create a replication connection that's authenticated by certificate
and relies on a username map to map from the CN in the client
certificate to the database username that's mentioned in the "hostssl
replication" entry.
All I can suggest at this point is to go over your configuration
with a fine-tooth comb, looking for probably-silly mistakes such as
inconsistent spellings. One thing I can think of to mention in
particular is to be sure that the standby's primary_conninfo
explicitly includes "user=pgrepmgr_nonprod", as that's likely not the
user name it'd default to.
Another idea could be to enable log_connections on the primary,
and see if the incoming connection request looks different than
you were expecting.
regards, tom lane
^ permalink raw reply [nested|flat] 2+ messages in thread
end of thread, other threads:[~2024-06-21 18:24 UTC | newest]
Thread overview: 2+ messages (download: mbox mbox.gz follow: Atom feed)
-- links below jump to the message on this page --
2024-06-21 17:21 Re: Replication using mTLS issue Drew Zoellner <[email protected]>
2024-06-21 18:24 ` Tom Lane <[email protected]>
This inbox is served by agora; see mirroring instructions
for how to clone and mirror all data and code used for this inbox