maria-developers team mailing list archive

[Jeremy Zawodny <Jeremy@xxxxxxxxxxx>]

On Mon, Jan 25, 2010 at 11:47 AM, Alex Yurchenko <
alexey.yurchenko@xxxxxxxxxxxxx> wrote:

> On Mon, 25 Jan 2010 10:47:23 -0800, Jeremy Zawodny <Jeremy@xxxxxxxxxxx>
> wrote:
> > If the connection between a slave and master is interrupted, the slave
> > won't
> > report itself as being "behind" until the slave's network timeout to the
> > master expires and it reconnect (assuming it can).
> >
> > Jeremy
>
> Hi Jeremy,
>
> Does it mean that you want to detect network outages using the heartbeat?
> If so, won't it be easier to just set master-slave connection timeout to
> the desirable value and let replication system to handle it? I'm asking
> this not from position of improving current replication in short term, but
> from position of whether a heartbeat function is needed in replication API?
>

Not at all.

My point is that you cannot (today) rely on MySQL slaves to "self report" in
all situations.  With a heatbeat that the master sends every N seconds and
the slave expects to see every N seconds, it's possible to see latency in
the replication stream regardless of the cause--even during weird network
outages that the slave won't notice until the timeout is reached.

Does that make more sense?

Jeremy