On Dec 6, 2007, at 7:30 PM, Radu Handorean wrote:
Hi,
I'm new to Linux HA and I am already a bit puzzled by a few things.
I have two AMD64 Opteron machines (same hardware config) running
SuSE 10.2 and HA. I managed to configure HA using the GUI and all
the wizards I managed to bump into. It seems to work. The thing is I
look at it like at a blob of something unknown to me and I keep
circling it and poking it trying to understand how to interact with
it...
First, I am a little confused by some terminology/behavior. It seems
I can use the GUI on each of my two nodes to set them to standby or
active mode (each of the 4 combinations is accepted). I also read
that a node can be master, DC (decision coordinator), active,
passive. Well, all of a sudden, a node can be many things I think i
understand individually but all together confuse me. In my mind one
node in the (2-node) cluster (in my case) is the boss (active) and
the other one is in standby waiting to become active after a
failover. This simple model I have in my mind is definitely
challenged by the plethora or adjectives I listed above, plus by
combinations of all active (everyone is the/a boss) / all standby
(anarchy) which seems to be acceptable.
the DC is the boss but that has no impact on where resources get run.
its just who decides where they all go.
In addition to this, I don't really understand how to interact with
HA once it is up and running. I am writing some software that needs
to be able to query the cluster and find out information about the
nodes. It seems crm_mon can help with this. It gives the list of
nodes and which one is DC ( is this what I think of as THE active
node / the boss?). I get the host names and I can figure their IP
addresses separately so that should be fine (wouldn't mind if I
could find this info from HA directly, but doing this separately is
not a big deal).
have a look at the code in crm_mon, its not hugely complex.
Another source of confusion for me is the fact taht using the GUI I
need to connect to a cluster and provide a password while invoking
crm_mon from the command line goes straight to a"default" cluster
(the only one actually) and fetches information without any need for
a password. Now I understand that the GUI does more than return some
info, it can actually manipulate the cluster but would this be the
reason why I don't need a password for crm_mon? It feels like
anybody should be able to look and one's bank account but it takes a
password to make a payment or transfer funds. The first half seems
just wrong.
the gui can run on any machine
crm_mon can only be run from a cluster node as certain users.... so in
effect you provided the password when you logged in
I would also like to be able to force a failover and I am guessing
there should be some command that would do this, is this true? How
do I force a failover? Is there a way to figure that the failover is
complete?
crm_standby or crm_resource -M depending on exactly what you're trying
to achieve.
Sorry about the avalanche of blurb and questions, but I really hit
my head against all of these and I tried to answer them by myself
before asking for help here and they kind of piled up on me.
Thanks for any kind answer you may have for me! Pointers to
literature would be greatly appreciated (I have read the linux-
ha.org webpages as well as some Novell documentation and yet I had
to send this email out).
____________________________________________________________________________________
Be a better friend, newshound, and
know-it-all with Yahoo! Mobile. Try it now.
http://mobile.yahoo.com/;_ylt=Ahu06i62sR8HDtDypao8Wcj9tAcJ
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems
