On Thu, Aug 12, 2010 at 01:31:28AM +0200, Dejan Muhamedagic wrote:
> On Wed, Aug 11, 2010 at 03:59:34PM -0700, David Lang wrote:
> > On Thu, 12 Aug 2010, Dejan Muhamedagic wrote:
> > > On Wed, Aug 11, 2010 at 02:44:36PM -0700, David Lang wrote:
> > >>> haresources2cib.py is obsolete and probably produces a bad
> > >>> cib.xml. The recommended way is to create a configuration using
> > >>> the crm shell.
> > >>
> > >> Ok, so this means that there is officially no migration path for those 
> > >> of us
> > >> using a V1 sty;e config
> > >
> > > haresources2cib had value once the XML was the only way to
> > > configure CRM. I think that converting haresources by hand
> > > doesn't take much effort anymore. It is also a good way to get
> > > acquainted with pacemaker and the crm shell. That was the reason
> > > haresource2cib was retired.
> > 
> > opeque config files that can only be edited with special tools are the bane 
> > of 
> > sysadmins who need to manage lots of boxes.
> 
> Yes. But crm shell is not a special tool, such as tools for
> windows registry or aix objdb. The idea of the shell language is
> to have a 1-1 translation to XML and back.
> 
> > I currently manage over a hundred 
> > clusters of machines. with v1 style configs this is easy to integrate into 
> > the 
> > other server management tools, if changes had to be done strictly via the 
> > crm 
> > shell, this is much more complicated.
> 
> Why would that be? If you do
> 
> crm configure edit
> 
> it will take you straight to the editor and the saved changes are
> going to be applied to the cluster. There's also (in v1.1)
> 
> crm configure filter
> 
> which can be used with say sed. There's also a way to load/save
> configurations from/to regular files.
> 
> As for the management, how do you make a node standby now?
> hb_standby, right? How's that different from "crm node standby"?
> 
> > >> This is really starting to sound like we need to fork heartbeat back to 
> > >> the
> > >> 2.x or thereabouts when it could work for simple things easily.

Nonsense.

You just use heartbeat as is.

Ok, so there has been a package split.  You don't install one package
anymore, you installe a few extra dependencies. Get over it.

Your apt or rpm or yum or zypper or whatever will resolve those
dependencies for you anyways. If the do not, complain to the responsible
packagers for your distro.

And then you use haresources mode, if you insist.
But skip down to, and follow, the configuring-heartbeat-v1-was-so-simple
link below, anyways.  Please.

And if that does not work now for you as you would like it to, it very
likely had done the same in heartbeat 2.1.x or 1.2.y.

If you can prove an actual regression in working with haresources mode,
something that works for you with older heartneat, but does no longer
work like that with heartbeat 3, all other things being equal,
report that here. We'll track it down.

Only if it happens to be something that "has been like this always",
it is unlikely that someone steps forward to fix it, now.

> > most of us who didn't need that complexity just ignored it as long as the 
> > haresources configs continued to work.
> 
> And, for the time being, they should work.

To the best of my knowledge, if you just want haresources, you install
heartbeat 3 and its dependencies (cluster glue, agents), and be as happy
or unhappy as you always was with haresources style configs.

Any bug or strange behaviour you may encounter is unlikely to be a
regression, I'd suggest "it has always been that way" until proven
otherwise. Especially for these strange "no resources for simultaneous
reboot": some blame "the new heartbeat" or even pacemaker for this
"intransparent behaviour".
I think if you use 2.1.4 with the same resource scripts, config and
haresources, it will happen just the same.

No I cannot say what is wrong there from the logs.  These logs have been
changing, and usually did not match the description accompanying them in
the mail.  One time it was an error like "ERROR: Cannot locate resource
script ", and some was complaining that it should at least say _what_
script it could not locate. Well, it DOES.

The "resource manager" of haresources mode is a shell script.
It usually can be found in /usr/shared/heartbeat/ResourceManager,
gets some environment variables, and arguments, and that's it.
The line in that script is
   ha_log "ERROR: Cannot locate resource script $script"

So either the logs presented have been mangled, or $script seems to have
been empty. I've seen this if people manage to get "strange characters"
into haresources, like "unbreakable space" or something else that is
"invisible" on the terminal but non-whitespace to the shell.

Then, at other times, it was a full list of resources that was actually
taken over. Which looked ok. But apparently was not, at least according
to the poster. I'm not sure what the expected behaviour was at that time.

Also (and now I'm speculating heavily) the old heartbeat changelog says
there have been bugs in simultaneous reboots before (a long time ago),
related to autofailback on. Maybe these have been races, and not been
fixed but only reduced in likelyhood; I don't know.
Try with autofailback off, which would be the recommended setting for an
"both nodes are equal" setup, anyways.

> Don't know what will the future bring, didn't notice much interest in
> supporting that.  Perhaps somebody from Linbit can comment too.

We (LINBIT) certainly help anyone to convert v1 haresources style
clusters to pacemaker clusters. We usually don't actively support
haresources style clusters for new customers anymore, though we still
maintain a few such installations.

It is sufficiently easy to convert an haresources to a pacemaker config
using the crm shell. For a nice read on that, see
http://theclusterguy.clusterlabs.org/post/178680309/configuring-heartbeat-v1-was-so-simple

heartbeat + pacemaker makes a nice cluster.
Or corosync + pacemaker, if you have to ;-)



For those picking this up from the archives:
This thread was in part about how a linux admin loves to
edit simple text files with his favorite editor
(and how they hate to do that with xml files; but you don't need to do
that anymore, you can do pacemaker without xml just fine).

The CRM shell allows you to edit simple text files:
http://www.clusterlabs.org/doc/crm_cli.html
crm configure edit (or show, or load, or save, ...),
or even filter it through sed if you so chose.
It even allows you to do that in a "shadow" config, to do a dry run and
see what the cluster would do, if you had changed the live config that
way. Of course you only need that for some more interessting stuff than
just adding an other IPaddr instance.

In case you are more the GUI kind, have a look at the DMC,
http://www.drbd.org/mc/management-console/
At the time of writing, the latest release announcement is
http://www.gossamer-threads.com/lists/drbd/announce/19904


-- 
: Lars Ellenberg
: LINBIT | Your Way to High Availability
: DRBD/HA support and consulting http://www.linbit.com

DRBD® and LINBIT® are registered trademarks of LINBIT, Austria.
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

Reply via email to