communication #138311
openmx1 - relayhost set to relay.i.o.o ?
0%
Description
I noticed that mx1 and mx2 were not handling mails the same way - mx2 sends e.g. progressoo.i.o.o directly to progressoo, as it should, but mx1 sends it via anna (relay.i.o.o). I was just wondering if anyone knew who might have changed that setting on mx1? I don't recall doing it myself and it is wrong.
Updated by cboltz 6 months ago
If I had to guess...
mx1 (mx1.o.o):/home/cboltz # testhighstate
local:
----------
ID: /etc/postfix/main.cf_relayhost
Function: file.replace
Name: /etc/postfix/main.cf
Result: None
Comment: Changes would have been made
Started: 17:08:09.293830
Duration: 55.396 ms
Changes:
----------
diff:
---
+++
@@ -712,7 +712,7 @@
mynetworks_style = subnet
# poo#138311
#relayhost = [relay.infra.opensuse.org]
-relayhost =
+relayhost = [relay.infra.opensuse.org]
alias_maps =
canonical_maps =
----------
[...]
therefore I'd guess it was caused by a highstate run - mx1 doesn't have any roles and therefore gets the default config with the relayhost.
Which reminds me that we should integrate the few things that are missing from the mailserver role, and apply that role on all mx* servers.
mx2 already has the mailserver role, so running testhighstate there will show what needs to be updated in salt.
Per, can you please check testhighstate output on mx2 and either integrate it in salt, or at least confirm that all diffs are intentional so that someone else can copy them over to salt?
Updated by pjessen 6 months ago
- Assignee set to pjessen
cboltz wrote in #note-3:
therefore I'd guess it was caused by a highstate run - mx1 doesn't have any roles and therefore gets the default config with the relayhost.
Two things that surprise me -
1) who runs a highstate only on mx1?
2) I thought Olav and I had "fixed" the MX config in salt. (maybe my memory is at fault). We were preparing running up an MX3 and MX4 during the datacentre move.
Which reminds me that we should integrate the few things that are missing from the
mailserverrole, and apply that role on all mx* servers.
mx2 already has the mailserver role, so running testhighstate there will show what needs to be updated in salt.
Ah, maybe that is the problem - that mx2 is actually correct and in place, whereas mx1 is not.
Per, can you please check
testhighstateoutput on mx2 and either integrate it in salt, or at least confirm that all diffs are intentional
so that someone else can copy them over to salt?
Willdo.
Updated by cboltz 6 months ago
pjessen wrote in #note-4:
1) who runs a highstate only on mx1?
Nobody ;-) Actually Georg applied role.base everywhere some day ago, and - after finding out that postfix is part of role.base - manually reverted the postfix changes on mx2. The change on mx1 was probably small enough to be overlooked ;-)
2) I thought Olav and I had "fixed" the MX config in salt. (maybe my memory is at fault). We were preparing running up an MX3 and MX4 during the datacentre move.
You indeed had, but since then did some manual changes (maybe some of them even "only" for testing) to the postfix config.
Getting these small changes into salt shouldn't be too hard ;-)