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

Re: [libvirt] [PATCH v2 01/12] Define keepalive protocol



On Thu, Sep 29, 2011 at 03:46:05PM +0200, Jiri Denemark wrote:
> On Thu, Sep 29, 2011 at 14:19:43 +0100, Daniel P. Berrange wrote:
> > On Fri, Sep 23, 2011 at 10:24:46AM +0200, Jiri Denemark wrote:
> > > The keepalive program has three procedures: ADVERTISE, PING, and PONG.
> > > All are used only in asynchronous messages and the sender doesn't wait
> > > for any reply. However, the party which receives PING messages is
> > > supposed to react by sending PONG message the other party, but no
> > > explicit binding between PING and PONG messages is made. ADVERTISE is
> > > sent by a client to indicate it supports keepalive protocol. Server is
> > > not allowed to send any keepalive message until it sees ADVERTISE.
> > 
> > I guess I'm not entirely understanding what the point of the
> > ADVERTISE message here is?
> > 
> > IIUC, the flow of messages you are describing will end up as:
> > 
> >  1. C -> S  remote_supports_feature_args (KEEPALIVE)
> >  2. S -> C  remote_supports_feature_ret  (TRUE|FALSE)
> >  3. C -> S  keepalive ADVERTISE
> >  4. C -> S  keepalive PING
> >  5. S -> C  keepalive PONG
> >  6. C -> S  keepalive PING
> >  7. S -> C  keepalive PONG
> >  ...
> >  n. C -> S  keepalive PING
> >  n+1. S -> C  keepalive PONG
> >
> > We need to the remote_supports_feature method to determine if the
> > keepalive protocol is supported, what purpose is the ADVERTISE
> > message serving ?
> 
> PING messages can be sent by both client and server and sending them can be
> independently disabled on both sides. ADVERTISE is sent because a server may
> never get any PING messages from a client which was configured not to send
> them. But the server still wants to know that the client supports this feature
> and thus the server can send PING messages. Also there is a keepalive timeout
> between ADVERTISE and PING, which can be different on both sides so server may
> decide to send PING earlier than the client.

Ok, in the case where the intervals are different it appears that
whomever has the shorter interval will start PING'ing first, and
thus the other side will only ever send PONGs, and never try to
send PINGs itself. So we're not getting any duplicated PING/PONG
in both direction which is good.

> The server could probably deduce that the client supports keepalive from the
> fact that the client calls remote_supports_feature(KEEPALIVE) but that seems
> weak to me and sending a separate ADVERTISE message looks better :-)

I've got some comments about initialization, but I'll post them against
another patch.

Daniel
-- 
|: http://berrange.com      -o-    http://www.flickr.com/photos/dberrange/ :|
|: http://libvirt.org              -o-             http://virt-manager.org :|
|: http://autobuild.org       -o-         http://search.cpan.org/~danberr/ :|
|: http://entangle-photo.org       -o-       http://live.gnome.org/gtk-vnc :|


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