Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]


Groups > comp.protocols.time.ntp > #164241

Re: best current practices regarding "peer"?

From "Miroslav Lichvar via questions Mailing List" <questions@lists.ntp.org>
Newsgroups comp.protocols.time.ntp
Subject Re: best current practices regarding "peer"?
Date 2026-03-23 11:53 +0000
Organization Taughannock Networks, Trumansburg NY
Message-ID <acEop7FmCfxBZjYa@localhost> (permalink)
References <OPrK_ibu858KdMgqHuQYFfB0TH31AV-x@violet.siamics.net>

Show all headers | View raw


On Thu, Mar 19, 2026 at 01:25:23PM +0000, Ivan Shmakov wrote:
> 	To my surprise, I’ve found that as part of the ‘hardening’ effort
> 	for NTPsec (the version included in Debian Trixie) the ‘peer’
> 	association type got removed; this is what I get in syslog:
> 
> CONFIG: peer deprecated, treated as server: (peer-IP-address)
> 
> 	The question is: what is the best current practice for having 
> 	‘a clique of low-stratum peers operate as mutual backups for each
> 	other’?  Is ‘peer’ truly deprecated for any and all possible uses?

"peer" configures a symmetric association. "server" configures a
client-server association. A symmetric association can be fully
replaced by two client-server associations in the opposite directions.
It just doubles the network traffic. Basically, the client-server mode
is a special case of the symmetric mode. NTP originally had only
the symmetric mode, later was added client-server mode, and couple
decades later the symmetric mode might be removed from NTP completely.

The current best practice for "a clique of low-stratum peers operate
as mutual backups for each other" is the same except use "server"
instead of "peer".

The only interesting feature of the symmetric mode is that time can be
pushed to a host that doesn't specify the other peer in its config.
The disadvantages are much more difficult implementation, worse
security, and less accurate measurement of the network delay due to
much larger intervals between requests and responses.

-- 
Miroslav Lichvar

Back to comp.protocols.time.ntp | Previous | NextPrevious in thread | Find similar


Thread

best current practices regarding "peer"? "Ivan Shmakov" <ivan@siamics.net> - 2026-03-19 15:18 +0000
  Re: best current practices regarding "peer"? "Harlan Stenn via questions Mailing List" <questions@lists.ntp.org> - 2026-03-21 00:53 +0000
  Re: best current practices regarding "peer"? "Miroslav Lichvar via questions Mailing List" <questions@lists.ntp.org> - 2026-03-23 11:53 +0000

csiph-web