[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: CARP using Serial to sync
- To: misc_(_at_)_openbsd_(_dot_)_org
- Subject: Re: CARP using Serial to sync
- From: David Terrell <dbt_(_at_)_develest_(_dot_)_com>
- Date: Tue, 11 May 2004 17:17:29 -0500
- Mail-followup-to: David Terrell <dbt_(_at_)_develest_(_dot_)_com>, misc_(_at_)_openbsd_(_dot_)_org
- Reply-to: David Terrell <dbt_(_at_)_develest_(_dot_)_com>
On Tue, May 11, 2004 at 03:58:33PM -0400, Chuck Yerkes wrote:
> There's an internal 10 pin header that is the second serial port.
> When I did lots and lots of Failover/HA machines we mandated dedicated
> heartbeat interfaces (on separate cards). This was because the
> Suns COULD get lots and lots of traffic (before network switches
> existed, even) and a heartbeat could get lost. We dropped SL/IP
> support when Sun's slip stuff proved to be unreliable.
> Veritas can use FC/AL and SCSI for heartbeat. Since a pair shared
> disk ANYWAY, this makes terrific sense. And all it needed was the
> "little sister mode" of "hey, you alive?" and response.
This is a really cute idea but even their own techs say "don't use a
shared disk for heartbeat, nobody does that"
General technique is doing heartbeat across multiple links, preferably
separate cards (not two ports on a quad card), preferably on different
backplanes on bigger iron machines.
Visit your host, monkey.org