Information Technology Reference
In-Depth Information
In preparation
: The partially con
gured concrete workflow is imported from
the gUSE application repository using the ASM function
importWork
fl
ow
. This
function returns a workflow ID that is used to monitor the workflow status. This
ID is attached to the submission entity and saved in eCAT. Then the required
inputs for that particular submission are transferred to the grid data resources
using the DT component. After that, the workflow is con
gured completely by
setting the input and output URIs, which de
ne the location to retrieve inputs
and save outputs on the grid data resources. This is done using the ASM
functions
setRemoteInputPath
and
setRemoteOutputPath
, respectively, for all
input and output ports. At the end, the con
gured workflow is submitted to the
gUSE WFI by using the ASM function
submit
. For optimization purposes, these
steps are performed asynchronously for each submission within a separate
thread.
In progress
: The gUSE WFI executes the workflow by submitting the jobs to the
compute resources. The DCI Bridge is utilized to communicate with various
compute resources. The DCI Bridge is con
gured to automatically resubmit one
time whenever a job submission to the gLite WMS fails. Additionally, the DCI
Bridge is con
gured to include the following by default in the job description
language (JDL)
cations with the
gLite WMS: (a) the address of the MyProxy server to enable automatic renewal
of community grid proxy at the grid job level, which is particularly important for
long running jobs; and (b) environment variables that are required for proper set-
up and utilization of CernVM-FS server.
files that are used to communicate the job speci
Transfer outputs
: The outputs of the workflow execution are transferred from
the grid data resources to the data server.
On hold
: If any problem occurs during any of the previous states, the submission
moves to this state, and a noti
cation is sent to the system administrators. An
administrator then investigates and troubleshoots the error using information
about that particular submission that is presented on the administrator
'
s dash-
board (see Fig.
10.4
). If the error is
fixed, the workflow is resumed, and gUSE
continues execution from the last successful job. This is often the case for
transient errors related to the DCI, for example, because of a terminated job or
an unavailable
le.
Failed
: If the problem cannot be solved, the administrator aborts the submission,
which causes it to move into this state. At this point a message can be written by
the gateway administrator to the user providing high-level information about the
cause of the error and the actions to take. Typically, these errors occur due to
invalid inputs or application errors, as all DCI-related errors are handled auto-
matically by gUSE or manually by the administrator.
Done
: If all actions are successful the end-user gets a noti
cation and can
download the outputs via the web interface.
Search WWH ::
Custom Search