CCL: Restart CCSD jobs
- From: Roger Robinson <r.robinson#imperial.ac.uk>
- Subject: CCL: Restart CCSD jobs
- Date: Mon, 22 Oct 2012 16:08:44 +0100
Sent to CCL by: Roger Robinson [r.robinson : imperial.ac.uk]
Hi,
I need to restart some CCSD jobs as the machine they are on are being
decommissioned.
The jobs have produced some very large rwf ie between 100GB and 270GB.
I have copied the rwf files back from local sites to the network device.
I notice the RWF havent been rewritten to for several days.
They have all got to the part of the job after the convergence
iterations ie all say something like this.
Iteration Nr. 14
**********************
DD1Dir will call FoFDir 1 times, MxPair= 600
NAB= 300 NAA= 0 NBB= 0 NumPrc= 16.
DE(Corr)= -1.5338361 E(CORR)= -349.45990574 Delta=-1.89D-08
NORM(A)= 0.12459143D+01
Largest amplitude= 3.48D-02
The scratch files have also stopped growing.
Is there still alot of input and output to the scratch files anymore
after this point ?
Would it be possible to carry out the jobs with the scratch file on
the
network drive at this point ? I should I move then to new local file
systems ?
As a test I restarted one job with the restart file on the network
file
system. The job seems to be running ok and more importantly is using
100% of the CPU allocated which could suggest its not struggling for
I/O
throughtput ?
Thanks Roger