After submitting a job to our pool (v 6.7.14) I asked it to give me a summary of which VMs on that machine were currently in use. To my surprise I got differing answers depending on which VM was being queried.
condor_status -format "Name: %s\t" Name -format "vm1_State: %s\t" vm1_State -format "vm2_State: %s\t" vm2_State -format "vm3_State: %s\n" vm3_State
Name: vm1@app13 vm1_State: Claimed vm2_State: Unclaimed vm3_State: Unclaimed Name: vm2@app13 vm1_State: Unclaimed vm2_State: Unclaimed vm3_State: Unclaimed Name: vm3@app13 vm1_State: Unclaimed vm2_State: Unclaimed vm3_State: Unclaimed
After 4 minutes this fixed itself and I got the expected output
Name: vm1@app13 vm1_State: Claimed vm2_State: Unclaimed vm3_State: Unclaimed Name: vm2@app13 vm1_State: Claimed vm2_State: Unclaimed vm3_State: Unclaimed Name: vm3@app13 vm1_State: Claimed vm2_State: Unclaimed vm3_State: Unclaimed
Is this supposed to take this long? I'd like to use vm_State to encourage jobs to distribute themselves breadthwise before going in deep on a single machine, but if these attributes only update every 4 minutes then jobs that come in close together are going to cause me lots of problems.
Thanks in advance for any help -Colin Little Fair Isaac Corp |