On 8/15/07, Michael Schwartzkopff <[EMAIL PROTECTED]> wrote:
> Am Mittwoch, 15. August 2007 10:52 schrieb Andrew Beekhof:
> > On 8/15/07, Christian Rishøj <[EMAIL PROTECTED]> wrote:
> > > On 8/14/07, Michael Schwartzkopff <[EMAIL PROTECTED]> wrote:
> > > > is it possbile to monitor the CPU load constantly and to move resources
> > > > to another node or declare on node dead if the load exceeds some level?
> > > > Or if the HDD space is too low.
> > > > How can this be done?
> > > > Can I use the Sysinfo RA for this purposes?
> > >
> > > You could have a cronjob which would check the system load, translate
> > > it to some score and update an attribute in the CIB accordingly. Then
> > > use this score in some place constraint (in combination with a value
> > > for resource_stickiness), letting resources move away from their
> > > current node if the load score is above some threshold.
> > >
> > > For updating attributes in the CIB from a script, use this command:
> > >
> > >    attrd_updater -n your_attribute -v some_value
> >
> > spot on
> > the other suggestion of using monit would also work
>
> Hi,
>
> I would like to stick to sysinfo since this is in the frame of heartbeat. I
> have a cloe resource of sysinfo so that on every node one resource is
> running. The sib section for one node is:
>
>            <nvpair
> id="status-c624d6ab-eef3-413e-902d-89f53655afc7-probe_complete"
> name="probe_complete" value="true"/>
>            <nvpair id="status-c624d6ab-eef3-413e-902d-89f53655afc7-arch"
> name="arch" value="i686"/>
>            <nvpair id="status-c624d6ab-eef3-413e-902d-89f53655afc7-os"
> name="os" value="Linux-2.6.18-4-xen-686"/>
>            <nvpair id="status-c624d6ab-eef3-413e-902d-89f53655afc7-free_swap"
> name="free_swap" value="150"/>
>            <nvpair id="status-c624d6ab-eef3-413e-902d-89f53655afc7-cpu_info"
> name="cpu_info" value="Intel(R) Celeron(R) CPU 1.70GHz"/>
>            <nvpair id="status-c624d6ab-eef3-413e-902d-89f53655afc7-cpu_speed"
> name="cpu_speed" value="4303.76"/>
>            <nvpair id="status-c624d6ab-eef3-413e-902d-89f53655afc7-cpu_cores"
> name="cpu_cores" value="1"/>
>            <nvpair id="status-c624d6ab-eef3-413e-902d-89f53655afc7-cpu_load"
> name="cpu_load" value="0"/>
>            <nvpair id="status-c624d6ab-eef3-413e-902d-89f53655afc7-ram_total"
> name="ram_total" value="150"/>
>            <nvpair id="status-c624d6ab-eef3-413e-902d-89f53655afc7-ram_free"
> name="ram_free" value="50"/>
>            <nvpair id="status-c624d6ab-eef3-413e-902d-89f53655afc7-root_free"
> name="root_free" value="1.1"/>
>
> Yes, id did configure with the gui ...
>
> How can I do a placement constraint so that the webserver is running on the
> node that has a cpu_load <= 2 (or root_free >=1 [GB])? Any hint what i should
> put into the constraints section?
>
>  <constraints>
>    <rsc_location id="place_apache" rsc="sysinfo">
>      <rule id="prefered_place_apache" score="100">
>        <expression attribute="root_free"
> id="dcb4c245-ed75-491b-b6d1-6ad46a0aea4e" operation="gte" value="1"/>
>      </rule>
>    </rsc_location>
>  </constraints>
>
> does not seem to do the job. How to check an attribute for a specific host?

a specific host?

any host matching that condition will have 100 added to its original score.
depending on what other constraints you have (and which nodes have <
1Gb free) this may not have any effect yet.
_______________________________________________
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