[Linux-ha-dev] ocf script for mysql

[Alan Robertson]

Joe Bill wrote:
>> But, if you don't have all the hosts in your cluster
> which you want to
>> run the same service on installed in the same places,
> then you're on the
>> road to madness, and service outages due to human
> errors.
> 
> LOL :-)
> 
> Are you suggesting that OCF will restrict the number
> of versions of one service software it can run
> simultaneously on the same cluster to only one ?

To be a little more precise, OCF doesn't have any view of configuring
resources differently on different machines - one way or the other.  It
only has a view from the current machine.  However the CIB has a
cluster-wide view of what the parameters to a resource agent are at any
given point in time.

And notice the difference between what I said, and what you asked...

You asked for two versions to be installed on the same machine
simultaneously, and to automatically pick which one to use according to
some unspecified criteria - which changes according to how far you are
through the upgrade.

In a cluster, the way one normally does rolling upgrades doesn't require
this, at least not that I've heard of anyone doing.

What one normally does in a cluster would be to do rolling upgrades
_instead of_ what you asked.

For example, let's say you want to upgrade from Apache 2.0.3 to Apache
2.0.7.

Then you put a machine in standby mode, and upgrade apache to 2.0.7 on
that machine.  Then you bring it back out of standby, and force Apache
to migrate to that machine.  If you like it, you keep it.  If you don't,
you migrate it back and downgrade the installed version.  Assuming you
eventually get a version you like, you upgrade the other nodes in your
cluster in a similar fashion.  If you never get a good version, then you
go find another job ;-)

Note that this does not require two versions of the software to be on
the same machine at the same time.  Since you, naturally, tested this
before cutting it over into production, you have few backouts to
perform, and you have a simpler life, and a more productive one.

This doesn't preclude writing a resource agent which can do what you
ask, but it would have to be built into each resource agent.

If you need two versions on your machine simultaneously for some other
reason (like one customer needs one, and the other needs another), then
there are ways of dealing with this - without quite as much madness as
the upgrade case.

> Or that no provision will be made in OCF to help run
> simultaneously different versions of one service
> software, on the same cluster, i.e. Apache v1.3 and
> v2.0, either in the same instance or in different
> instances ?
> I was counting on Linux-HA to support
> "rolling-upgrades", if not of the OS or Linux-HA, but
> at least of the service software, with the provision
> to "roll back" if necessary. It is hard to imagine how
> this can be done without supporting multiple versions
> of the same service software, even under different
> service names mind you, each version installed in it's
> own specific location to prevent the overwriting of
> the old version by the installation of the new.
> 
> May I kindly ask your input/comments/reactions on the
> matter ?


Was this of any help?

-- 
    Alan Robertson <alanr at unix.sh>

"Openness is the foundation and preservative of friendship...  Let me
claim from you at all times your undisguised opinions." - William
Wilberforce