On 2026-03-23 at 11:07:23 UTC-0400 (Mon, 23 Mar 2026 16:07:23 +0100)
Matus UHLAR - fantomas via Postfix-users <[email protected]>
is rumored to have said:
On 2026-03-23 at 10:36:42 UTC-0400 (Mon, 23 Mar 2026 15:36:42 +0100)
Matus UHLAR - fantomas via Postfix-users <[email protected]>
is rumored to have said:
With DNS round-robin your users will complain whene either one of IP
addresses is unreachable.
On 23.03.26 11:01, Bill Cole via Postfix-users wrote:
This is not such a huge issue when you do the "round-robin" with very
short (non-zero!) TTLs on the A records. It was worse some years ago
when some resolvers would sort answers, but I don't think anything
still does that.
Do you mean providing different DNS record when one of servers is
down?
No. I mean that the TTLs should be short enough that on attempts after
the first, the client will recheck DNS and likely get a different IP.
Some users may notice that a connection starts slow when the first
attempt hits a bad IP but as the connection will be almost certainly be
retried either automatically or manually, it should work well enough to
forestall complaints, as long as outages are infrequent and short.
That has been my experience, although it has been years since I had a
partial failure which would test behavior.
I mean, providing multiple IP addresses: clients that first try to
connect to address that is down will complain that the connection
takes too long.
Properly-designed clients will retry additional IPs of an A record. If a
MUA doesn't do that automatically (as web browsers will) a user will
almost always try again after an initial timeout.
--
Bill Cole
[email protected] or [email protected]
(AKA @[email protected] and many *@billmail.scconsult.com
addresses)
Please keep discussion mailing list replies *on-list*
Not Currently Available For Hire
_______________________________________________
Postfix-users mailing list -- [email protected]
To unsubscribe send an email to [email protected]
