We seem to have a strange situation on tahini with messages sitting in the
queue forever (7 days currently), despite the standard single retry rule with
cutoff at 4d. There's this retry record (the local part is irrelevant):
| tahini# exim_fixdb /var/spool/exim/ retry
| Modifying Exim hints database /var/spool/exim//db/retry
| > R:***CENSORED***@???:<exim-users-bounces@???>
| 24-Sep-2007 13:04:10
| 0 error number: -44 SMTP error from remote mail server after RCPT
TO:<***CENSORED***@???>: host nox.aegisng.com [24.39.184.234]: 451
Temporary failure, please try again
| 1 extra data: 13133
| 2 first failed: 18-Sep-2007 02:58:55
| 3 last try: 24-Sep-2007 13:04:10
| 4 next try: 24-Sep-2007 19:04:10
| 5 expired: yes
and a bunch of T: retry records, most involving polaris.aegisng.com, the
secondary MX for the domain:
| T:polaris.aegisng.com:24.97.166.174:1IXEB6-0008L9-Bf -46 13133 SMTP error
from remote mail server after end of data: host polaris.aegisng.com
[24.97.166.174]: 451 Temporary failure, please try again later.
| 17-Sep-2007 12:10:09 17-Sep-2007 16:11:23 17-Sep-2007 17:41:23
| --
| T:polaris.aegisng.com:24.97.166.174:1IZ8Go-00072P-JO -46 13133 SMTP error
from remote mail server after end of data: host polaris.aegisng.com
[24.97.166.174]: 451 Temporary failure, please try again later.
| 22-Sep-2007 18:19:24 22-Sep-2007 18:19:24 22-Sep-2007 18:34:24
| --
| T:polaris.aegisng.com:24.97.166.174:1IYnhl-0003Kj-5S -46 13133 SMTP error
from remote mail server after end of data: host polaris.aegisng.com
[24.97.166.174]: 451 Temporary failure, please try again later.
| 21-Sep-2007 20:23:04 21-Sep-2007 20:23:04 21-Sep-2007 20:38:04
| --
They are running a broken greylisting solution that I've contacted them about,
and the primary MX usually returns 451 after RCPT, while the secondary
returns 451 after end of data.
It seems that the R: record causes exim to skip delivery ("retry time not
reached") on queue runs, despite being past final cutoff. It also seems that
the R: record is updated periodically on initial delivery attempts (of new
messages). But since the secondary MX gets past DATA, the messages are kept
on the queue and a T: retry record written.
So there seem to be some inconsistencies here. Should a R: record be written
if not all MXes tempfail on or before RCPT? And shouldn't delivery fail
immediately if there is a R: record with expired:yes?
--
Magnus Holmgren holmgren@???
(No Cc of list mail needed, thanks)
