search

hep-gc / cloud-scheduler

Automatically boot VMs for your HTC jobs
http://cloudscheduler.org
Apache License 2.0
3 stars 0 forks source link

VM Keepalive not being honoured #407

Closed rptaylor closed 8 years ago

rptaylor commented 8 years ago

I have +VMKeepAlive = "90"

but the VM was only up for about 3 minutes:

2015-10-08 16:13:46,904 - DEBUG - Scheduler - Booting VM for job gcecs.heprc.uvic.ca#40365.7#1444346006 on: mouse 2015-10-08 16:13:46,905 - DEBUG - Scheduler - CLUSTER Name: mouse, Address: http://mouse01.heprc.uvic.ca:5000/v2.0, Type: OpenStackNative, VM slots: 5, Mem: [240000] 2015-10-08 16:13:46,906 - DEBUG - Scheduler - No defined security groups for job - trying default value from cloud_resources.conf 2015-10-08 16:13:46,907 - DEBUG - Scheduler - Using security group: ['default'] 2015-10-08 16:13:48,372 - INFO - Scheduler - Created VM cluster address http://mouse01.heprc.uvic.ca:5000/v2.0 name PandaCern 2015-10-08 16:14:43,393 - DEBUG - VMPoller - VM: 3b949531-db84-4251-b864-2f5ea4f26704 on mouse. Changed from Starting to Running.

2015-10-08 16:17:01,345 - INFO - VMDestroyCmd - Destroying VM: 3b949531-db84-4251-b864-2f5ea4f26704 Name: mouse-3dd102cd-6aec-4372-8de2-ed2abe166d15.heprc.uvic.ca Reason: VMType atlas-worker is no longer required.

mhpx commented 8 years ago

https://github.com/hep-gc/cloud-scheduler/blob/dev/cloud_scheduler#L1463 Might be cases where the idle_start was not set in time, will investigate.

mhpx commented 8 years ago

confirmed, idle_start not getting set.