[Pacemaker] Clone set members colocation

[Florian Haas]

On 2011-11-08 22:42, Yves Trudeau wrote:
> Hi,
>    I am currently working on a replication solution adding logic to the
> mysql RA and I need to be able to turn ON/OFF a virtual IP based on a
> node attribute (Florian suggestion).  In order to achieve this, I
> created a clone set of IPaddr2 RA and a location rule using the node
> attribute setting.  It works almost correctly except that clone members
> stick a bit too much and don't failback.  Here's my config:
> 
> node testvirtbox1 \
>         attributes IP="10.2.2.160"
> node testvirtbox2 \
>         attributes IP="10.2.2.161" \
>         attributes standby="off"
> node testvirtbox3 \
>         attributes IP="10.2.2.162" \
>         attributes standby="off"
> primitive reader_vip_1 ocf:heartbeat:IPaddr2 \
>         params ip="10.2.2.171" nic="eth0"

You're meaning to manage an IP range, right? If so, then you're missing
unique_clone_address="true" here.

I also don't quite understand what you're trying to do with these "IP"
attributes, but that doesn't look like it's relevant here.

> clone reader_vips reader_vip_1 \
>         meta globally-unique="true" clone-max="3" clone-node-max="3"
> notify="true" ordered="true" interleave="true"
> location No-reader-vip-loc reader_vips \
>         rule $id="No-reader-vip-rule" -inf: readerOK eq 0

That location constraint would not trigger on nodes where the "readerOK"
attribute is not defined at all. Is that intentional? If not, I'd put:

location No-reader-vip-loc reader_vips \
  rule $id="No-reader-vip-rule" -inf: \
    not_defined readerOK or readerOK eq 0

> colocation reader_vips_dislike_reader_vips -200: reader_vips reader_vips

What is _that_ meant to achieve?

> property $id="cib-bootstrap-options" \
>         dc-version="1.0.11-a15ead49e20f047e129882619ed075a65c1ebdfe" \
>         cluster-infrastructure="openais" \
>         expected-quorum-votes="3" \
>         stonith-enabled="false" \
>         no-quorum-policy="ignore"
> rsc_defaults $id="rsc-options" \
>         resource-stickiness="100"
> 
> 
> 
> - Upon startup, the reader_vips clone are well spread on all 3 nodes
> - Running "crm_attribute -N testvirtbox1 -l reboot --name readerOK -s
> mysql_replication -v '0'"  removes the clone set member from
> testvirtbox1 as expected and add it to testvirtbox2 or testvirtbox3
> - But... running "crm_attribute -N testvirtbox1 -l reboot --name
> readerOK -s mysql_replication -v '1'" does not make it come back, one of
> the other node still have 2 clone members

That I'd consider expected. As I understand it, your resource stickiness
keeps the clone instance where it is.

> Is there a way to make it come back normally or is there something I
> don't do correctly?

Try removing the resource stickiness score from rsc_defaults.

Cheers,
Florian


-- 
Need help with High Availability?
http://www.hastexo.com/now