On Tue, 2009-05-12 at 16:24 +0200, Andrew Beekhof wrote:
> Have you tried crm_resource -C yet?
> That will clear away any errors and tell the cluster its ok to try again.

Thanks, that was part of it but I eventually ended up starting from a
clean slate.

Here's a rough summary of what I did to fix and configure it:

Erase the existing CIB:

# cibadmin --cib_erase

now check the cluster, there should be no warnings since its clean:

# crm_verify -LV

Other useful commands:

# crmadmin --nodes

# crmadmin --dc_lookup

# crm_resource --list

If there are complaints about orphaned resources:

# crm_resource --cleanup -r <resource name> -H <node name>

One might have to do the above for each resource and node combo.

Now check that no warnings are produced.

# crm_verify -LV

Now configure the cluster using hb_gui:

2-nodes, foo and bar, Debian Lenny; 1 virtual ip; 1 email-alert

1. add the VIP resource

type = native
id = virtual_ip
type = IPaddr
attr
  ip = 1.2.3.4

2. add the alert email

type = native
id = email_alert
type = MailTo
attr
  email = [email protected]
  Subject = bar

3. add location contraints

type = location
id = pref_foo_email_alert
expression
  #uname eq foo
score = 100

type = location
id = pref_bar_email_alert
expression
  #uname eq bar
score = 0

type = location
id = pref_foo-virtual_ip
expression
  #uname eq foo
score = 100

type = location
id = pref_bar-virtual_ip
expression
  #uname eq bar
score = 0

4. start the resources

confirm the VIP is aliases to the NIC and that one can SSH to it.

5. check cleanliness with:

# crm_verify -LV


test the cluster

init position:
VIP on foo
email_alert on foo

a)

  - put foo into stand-by
  - VIP should migrate to bar
  - check alert email recieved
OK

b)

  - make foo active
  - VIP should migrate back to foo
  - check alert email recieved
OK


c)

  - make bar stand-by
  - nothing should happen to the VIP
  - make bar active

OK


###

There we go, hope it helps someone.

By the way, mgmtd listens on port tcp 5560. hb_gui connects to this
using TLS encryption so an ssh tunnel is not necessary. A shame that one
cannot choose the port though.

One out-standing question I have is that if I reboot foo, then the
resources will migrate to bar but when foo comes back up the resources
migrate back to foo. I did not expect this to happen since I have
"auto_failback off" in ha.cf. Is this because I have "crm on" so that it
ignores ha.cf?

In my production scenario this was actually OK and did not cause a
problem when it happened (because other services on foo were configured
correctly). However, I would not like the resources to failback
automatically, I would like to do so manually via hb_gui/CLI or at least
have a toggle for this behaviour. How do I do this? Any pointers
welcome.


Regards

Imran Chaudhry
Software Engineer

SmoothWall Ltd
1 John Charles Way
Leeds
LS12 6QA

Tel: +44 (0)870 1 999 500
Fax: +44 (0)870 1 991 399

SmoothWall Ltd. - http://www.smoothwall.net/

This email and any attachments transmitted with it are confidential to
the
intended recipient(s) and may not be communicated to any other person or
published by any means without the express permission of SmoothWall
Limited.

Any views expressed in this message are solely those of the author.

See: http://www.smoothwall.net/emailnotice.html for the full text of
this notice.


_______________________________________________
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