Re: [Condor-users] Persistent offline machines ads when using a third party power management tool?

[Ian Chesal <ichesal@xxxxxxxxxxxxxxxxxx>]

Hi Lukas,

Maybe. I'm not sure OFFLINE=True is ever being set. But I never considered that it could be getting set and the unset before hibernation kicked it.

I'll try extending HIBERNATION_WAIT_INTERVAL to see if it helps the situation.

Regards,

- Ian

---

Ian Chesal

Cycle Computing, LLC

Leader in Open Compute Solutions for Clouds, Servers, and Desktops

Enterprise Condor Support and Management Tools

http://www.cyclecomputing.com

http://www.cyclecloud.com

http://twitter.com/cyclecomputing

On Tuesday, 31 January, 2012 at 2:56 PM, slebodnik wrote:

Are symptoms similar like in ticket

https://condor-wiki.cs.wisc.edu/index.cgi/tktview?tn=2564

Regards,

Lukas

On Tue, 31 Jan 2012 13:39:26 -0500, Ian Chesal

<ichesal@xxxxxxxxxxxxxxxxxx> wrote:

It seems like it should work but in my 7.6.5 pool any machine that's

hibernated by third party (external to Condor) power management

software fails to end up with an OFFLINE=True attribute in the

machine

ad at the collector and subsequently disappears from my list of

machines so it cannot be woken up by Rooster.

Condor appears to know that it's being hibernated. I see the

following the MasterLog for the machine when the third party tool

starts the machine hibernation procedure:

01/26/12 09:17:13 PowerEventHander: Some driver/application is asking

if we can enter hibernation

01/26/12 09:17:15 PowerEventHander: Machine entering hibernation

Looking through the source it doesn't appear that the event handler

for this event does anything. There's no sign that it's updating the

machine ad to let the collector know it's going offline. When the

collector-side OfflineCollectorPlugin runs the ad is purged, not off

lined. If I set the offline attribute on the machine ad to true

before

hibernating the machine by hand everything works. Unfortunately I

don't seem to be able to run a script from the hibernation tool

that's

in use, so I can't (at least not without great difficulty) follow

this

approach in the third party tool.

Is it not possible to have this third party hibernation offline the

machine ad when the hibernate signal is trapped?

(This is on Windows BTW…)

Regards,

- Ian

---

Ian Chesal

Cycle Computing, LLC

Leader in Open Compute Solutions for Clouds, Servers, and Desktops

Enterprise Condor Support and Management Tools

http://www.cyclecomputing.com

http://www.cyclecloud.com

http://twitter.com/cyclecomputing