nmh-workers
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Nmh-workers] [OT] Strange DNS for cyrus.andrew.cmu.edu was: IMAP/nm


From: Robert Elz
Subject: Re: [Nmh-workers] [OT] Strange DNS for cyrus.andrew.cmu.edu was: IMAP/nmh, again
Date: Fri, 27 Oct 2017 10:34:39 +0700

    Date:        Thu, 26 Oct 2017 22:42:12 -0400
    From:        Ken Hornstein <address@hidden>
    Message-ID:  <address@hidden>

  | I suspect they wanted to provide high reliability and have clients fall
  | over to one of the other servers quickly.

That wouldn't work ... just having multiple addresses advertised should
achieve that as well as it can be done (regardless of their TTLs) - beyond
that the issue is how quickly the client TCP gives up on its connect attempt.

Once it fails, it should move on to the next addr - but if none are valid
because the TTL has expired, then (if working properly, most clients assume
that any TTL must be long enough for them to use the answers however they want)
the effect of such a short TTL would be that the client would need to query
again, get all the results again, and might very well attempt to contact the
failed client again.

Short TTLs are used when the client must not remember the addr for an extended
period - for the reason Valdis mentioned, or because of NAT (in my earlier
message - though 0 is often used in that case) - they are a DNS caching control
and have no influence on anything else.

The closest to "failure handling" that might be relevant is to allow the
DNS server to update the address list quickly in the case that one of the
addresses becomes unavailable (or a new one is to be added, but that is
usually less urgent.)

kre




reply via email to

[Prev in Thread] Current Thread [Next in Thread]