Mailing List Archive

PERL Aproach
One issue that seems to be of some concern is the ability of HA
solutions to operate across various hardware and OSs. I know this is
the Linux-HA project, but I am curious if any thought is actively gone
into this situation. I see it being more of a issue with Linux than
with other platforms, as many people will be moving to Linux as a low
cost solution, often to augment their already existing platforms.

We at LGS have began work on some PERL HA utils. Our goal is to
provide basic IP takeover ( much like fake ) and Heartbeat control (
like heartbeat ) while keeping it totally non system dependent. Its
not the answer for HA requiring high speed monitoring but we believe it
can be very useful in the majority of IP and Service failover.

I will post the scripts here as soon as we get them to a acceptable
operational level( they work now but we want to test them some more ).
They are lacking in respect to some features already in heartbeat, but
as it is PERL new features are easily added.

We will support as many features found in heartbeat as is possible
in PERL. We are still rather new at this HA stuff so we may be way off
in left field here. Please let me know what you think.


Ron Burton
Linux General Store
Atlanta, Ga.
Sr. Systems Analyst
(404)881-8090
PERL Aproach [ In reply to ]
Ron Burton wrote:
>
> One issue that seems to be of some concern is the ability of HA
> solutions to operate across various hardware and OSs. I know this is
> the Linux-HA project, but I am curious if any thought is actively gone
> into this situation. I see it being more of a issue with Linux than
> with other platforms, as many people will be moving to Linux as a low
> cost solution, often to augment their already existing platforms.

Heartbeat has been ported to FreeBSD (although there is a small patch that isn't
incorporated yet). I suspect it will port to other systems fairly easily also.
I thought seriously about using Perl at one point, but decided on C because of
many experiences over a long period of time of minor flakiness in Perl itself,
and the relative size and speed of Perl solutions. If there is even one memory
leak in the interpreter, then as it runs over time this will be unacceptable.
Finding and fixing such a leak would become our problem. So would documenting
every version that had this leak, and detecting whether or not any given
installation had the leak.

The "portability" problems in heartbeat are things which will not likely improve
in Perl: Things like adding an alias, detecting whether an alias is currently
active or not are things that change from system to system, and are not likely
to be covered by Perl's many excellent modules.

However, there are many things in Linux-HA which are not yet completed, and one
of the components which we are likely to use is already in Perl. For
diagnostics, we're likely going to use Mon, which is written in Perl. This
would have the advantage of adding to what is being done instead of duplicating
it.

-- Alan Robertson
alanr@bell-labs.com