[DRBD-user] speed of fail-over..

Marian Marinov mm at yuhu.biz
Fri Oct 17 23:28:04 CEST 2008

Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.


On Saturday 18 October 2008 00:10:13 Stefan Seifert wrote:
> On Friday, 17. October 2008, Little, Kevin wrote:
> > I searched the archives for this, but found nothing.
> >
> > I've seen mentioned
> > (http://fghaas.wordpress.com/2007/06/26/when-not-to-use-drbd/) that the
> > fail-over time for DRBD+Heartbeat is on the order of 20 seconds.
> >
> > First, is this accurate?
>
> The number highly depends on your setup and the applications you're
> running. As such it cannot be accurate. What it does show is the order of
> magnitude you can expect. Depending on your setup it can also be two
> seconds or even a minute.
>
> > Second, what portion of the 20 seconds is a
> > function of Heartbeat, what portion is a function of DRBD?
>
> DRBD itself depends just on your ping-int and ping-timeout and maybe the
> timeout value. You can tune them to your system and can even get down to
> about one second of fail over time. But don't overdo it. If your system is
> just under very high stress and answers late, you may push it to a fail
> over at a time, where it's not needed and possibly the worst timing. Be
> sensible.
>
> In my experience most of the time is not used for detecting the fail over
> condition, but simply for starting services on the stand by node. This
> could be minimized if the applications can cope with getting their data
> files at runtime and thus can be started at system boot.
>
> Regards,
> Stefan

Also you have to understand that this article is written mainly for 
active/passive setups.

In active/active setups(if your application can handle this setup) the 
failover time is far less.

Regards
Marian Marinov



More information about the drbd-user mailing list