On Sep 7, 2008, at 6:51 PM, Andrew McGill wrote:
On Sunday 07 September 2008 16:35:34 you wrote:
On Fri, Sep 5, 2008 at 07:08, Andrew McGill <[EMAIL PROTECTED]>
wrote:
On Thursday 04 September 2008 21:12:31 Andrew Beekhof wrote:
On Sep 4, 2008, at 2:39 PM, Paul Walsh wrote:
Lars Marowsky-Bree wrote:
Well, heartbeat doesn't require pacemaker. Pacemaker doesn't
require
heartbeat (it can also run on openAIS). But yes, they can
cooperate.
I take it that's because Heartbeat can still use a "v1" style
configuration (though the terms "v1" and "v2" may add to
the confusion - people tend to want to use the "latest/newest"
version so perhaps see "v1" as inferior)?
It _is_ inferior.
Its easier to configure because it doesn't do even half what v2
can.
V1 lets you define groups that run here or there... of course the
configuration can be mastered in a nanosecond.
Maybe "simple configuration" for non-crm clusters and "advanced
configuration" for those with a CRM (Pacemaker or
otherwise)? OK, so "simple" may be a bit lightweight, but using an
haresources file is a lot easier than getting to
grips with the XML! :)
Its not _that_ hard. 90% of it is cut&paste and changing a few
values.
You'd think we were requiring people to learn Latin or something.
But we've been making the XML easier to read anyway and Dejan has
just
completed an awesome CLI tool (totally devoid of XML) for 1.0 that
makes configuring the cluster a walk in the park.
This is sorely needed. Weeks later, I still have no working
version 2
configuration to replace this one-liner:
ubu drbddisk::r0 vgchange::vg0 isnsd iscsi-target
IPaddr::10.3.7.10/24/eth0
haresources2cib.py was too hard to run?
It was not so much that I had to figure out that it silently ignored
line
continuations,
Did you tell anyone? Or should we magically know you had a problem?
but more that the resulting configuration caused heartbeat to
spew messages about how its internal ping timed out because
heartbeat itself
was using up all the CPU time.
That has absolutely _nothing_ to do with the crm or with the contents
of cib.xml
Ping time outs are purely related to the underlying messaging layer -
the same layer that haresources based clusters use.
Disclaimer: I have not drunk the xml kool ade.
Neither have I. But there are some thing it _is_ good for.
Like mixing static and dynamic content :) I get the feeling that
folks use
xml solely to avoid having to write parsers and generators.
A good reason on its own - when will people learn to stop reinventing
the wheel?
But it also allows XSLT to be used to, for example, upgrade the config
automatically in future versions.... which is exactly what 1.0 does.
(And on this
topic, the idea of giving everthing an id is a glorified line
number, and
those transitions are "goto" statements - except I'm not sure how it
helps to
point this out.)
You have a better way to tell two objects apart, to allow them to be
referenced from elsewhere in the config and to delete/update them?
Seriously, XML isn't the holy grail but these objections are just silly.
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems
