heartbeats

Hi!

AFAIK there must be at least two heartbeats for VERITAS Cluster.
(Version 4.0, SOLARIS)

My question is, what happens if one heartbeat fails?
What will happen to my cluster and applications.

regrads
Heinz

Hi,

ok, I've read some things about Joepardy.

Please let me explain, if I understood it correct:

In Joepardy the cluster resources are properly working,
but no automatic failover is now possible, until the broken
line is repaired.

Correct?

Heinz

Why have 2 links ?
Redundancy !!


OK, what does that mean.

You are machine 1 and you are linked to machine 2 via 2 links. 1 link
dies. Is it because the other machine is dying ? or just because
something in the path to the other machine (like a hub or a switch or a
NIC card) has died ?

You would not know. So, if you're not sure what is really happening on
the other machine, would you trust it to run all your applications ?

Would you trust it enough to give you the correct information regarding
it's state ? Would you trust it when it says that there wil be no danger
if you mount the filesystems now (yeah, mount it on both machines and
you have corruption).

The other reason why VCS will not do anything while a machine is in
jeopardy, is the fact that VCS will not know the difference between the
last link dying and the machine dying. All we will know, is the fact
that we can not "talk" to the other machine anymore. So what if we were
in the middle of failing stuff over from that machine, are we sure it
has been done ?

VCS will always try not to mess up data. Thus, not mount filesystems on
2 (or more) machines, not to run applications on more than 1 machine,
etc..... Thus avoid the potential for so-called "split brain". So, this
part is very important.

Now, as an administrator, you can actually do something about it. You
can force things to happen. (See autoenabling). But, then you take
resposibility (not VCS)

There must always be 2 private links. You can add low-pri links (these
are non-dedicated links). The low-pri links will only carry heartbeats
not cluster information.



Hoe that helps

Heinz wrote:[color=blue]
> Hi,
>
> ok, I've read some things about Joepardy.
>
> Please let me explain, if I understood it correct:
>
> In Joepardy the cluster resources are properly working,
> but no automatic failover is now possible, until the broken
> line is repaired.
>
> Correct?
>
> Heinz[/color]

Hi ME,

thank you for your detailled response.

Could you agree with that:

"In Joepardy the cluster resources are properly working,
but no automatic failover is now possible, until the broken
line is repaired."

regrads
heinz

no automatic fail-over --> can still force it by hand

Heinz wrote:[color=blue]
> Hi ME,
>
> thank you for your detailled response.
>
> Could you agree with that:
>
> "In Joepardy the cluster resources are properly working,
> but no automatic failover is now possible, until the broken
> line is repaired."
>
> regrads
> heinz[/color]