[Pacemaker] Enable remote monitoring

[Dejan Muhamedagic]

On Mon, Nov 12, 2012 at 10:25:22PM +0100, Arnold Krille wrote:
> On Monday 12 November 2012 10:50:57 Dejan Muhamedagic wrote:
> > Hi Arnold,
> > 
> > On Sun, Nov 11, 2012 at 07:37:29PM +0100, Arnold Krille wrote:
> > > On Sun, 11 Nov 2012 18:37:04 +0100 Dejan Muhamedagic
> > > 
> > > <dejanmm at fastmail.fm> wrote:
> > > > On Fri, Nov 09, 2012 at 05:22:08PM +0100, Lars Marowsky-Bree wrote:
> > > > > On 2012-11-09T14:06:29, Dejan Muhamedagic <dejanmm at fastmail.fm>
> > > > > 
> > > > > wrote:
> > > > > > > And also doesn't really help with getting the state/readiness of
> > > > > > > services the guest might provide.
> > > > > > 
> > > > > > Isn't it that one gets a login prompt only once the host reached a
> > > > > > certain run level?
> > > > > 
> > > > > Services may start in the background. The console may either be
> > > > > text, graphical, or network only.
> > > > 
> > > > Hmm, I have yet to see a Linux/UNIX host without a console. And
> > > > that means quite some time ;-)
> > > 
> > > While (almost) all systems have some kind of console, not all systems
> > > stay at a text-console during runtime.
> > 
> > It seems like you are mixing things up. This is not about
> > watching a computer monitor. On Linux there's always a console
> > tty.
> 
> 1) There are linux systems without a tty.
> 2) Watching the first / whatever tty is not telling you anything. Some systems 
> don't run a login on tty1 but show tty1 after start, some systems switch to 
> graphics (yes, virtual machines do that too), some machines show a nice menu 
> letting you choose your language and stuff.
> 
> > > If you want to watch  a
> > > linux-terminal-server, its not a text-console that is the primary
> > > screen.
> > > 
> > > And then watching the text-console for a login:-prompt tells you that
> > > the machine is up. "Pressing" enter tells you that the login-process is
> > > still answering. But that doesn't tell you whether the webserver on
> > > that machine is still working correctly.
> > 
> > That is not the point.
> > 
> > > Its not even telling you if
> > > the machine is reacting to network stuff. All it tells you is that some
> > > kind of system is started. So its actually only a little more then
> > > "virsh list |grep <machinename>" tells you.
> > > 
> > > Imho watching the console for monitoring a machine is as useless as it
> > > can get.
> > 
> > It tells you that a host is at a certain runlevel. The host is
> > ready for business, i.e. all other service at the particular
> > runlevel have been started. Otherwise, it wouldn't make sense
> > that it offers a login prompt. I don't know how it is with
> > systemd, perhaps systemd functions differently. Whether a
> > particular service is functioning properly or at all is up to the
> > monitor resource to find out.
> 
> Watching the/a tty of a linux machine is not telling you anything in general. 
> It's only us humans that can make some sense of what we see there. For a 
> machine to watch another generic machine via tty its to much variables and to 
> much engineering needed.
> 
> BTW: Some machines present the login:-prompt (text and graphics) well before 
> everything is started, suse and ubuntu did that some three-five years ago 
> already.
> 
> And actually I don't care (cluster-wise) at which runlevel the webserver is, 
> it has to serve pages at port [80,443] otherwise its useless and has to be 
> killed and restartet. No monitoring of any tty can tell me that...

Is it that you cannot read or refuse to read? This is _not_ about
monitoring a webserver or whatever other service you have, it's
just about whether a VM can reasonably be considered to have
reached certain state. It is up to _other_ resources to monitor
your proverbial webserver. What a discussion.

Thanks,

Dejan

> Have fun,
> 
> Arnold



> _______________________________________________
> Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
> 
> Project Home: http://www.clusterlabs.org
> Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
> Bugs: http://bugs.clusterlabs.org