Skip to content

Latest commit

 

History

History
145 lines (98 loc) · 4.64 KB

QUICKSTART.md

File metadata and controls

145 lines (98 loc) · 4.64 KB
Raw

Quick start for SHREK

Cloning the repository and initial setup

First clone the repository and change into the shrek directory

$ git clone https://github.com/BNLNPPS/shrek
$ cd shrek

Next, you'll want to modify the site configuration file, using your favorite text editor

$ cp shrek/config/site.yaml .
$ edt site.yaml

For sPHENIX users, there are two fields which need to be modified in order to get the system running:

Shrek/submissionPrefix points to a directory where shrek will create

Shrek/submissionPrefix points to a directory where shrek will create the job submission documents for PanDA and gather any supporting files needed to run the job. You should create a fresh directory and initialize a local git repository inside of it. And make sure the site.yaml file points to it:

$ mkdir ../submit  # directory should match the submissionPrefix in site.yaml
$ cd ../submit
$ git init
$ cd -

Donkey/scope

This is the target scope used by the donkey application to discover datasets. It is typically user.${USER}

Running tests

Next, you should run the test suite to confirm codes are setup properly.

$ pytest -v tests/

Setting up handy aliases

$ alias donkey='shrek/scripts/watchRucioForNewDatasets.py'
$ alias shrek='shrek/scripts/submitWorflowToPanDA.py'

Running test jobs with shrek

Issuing the following command will run the shrek processor on the simple chain workflow, creating the directory "sP00a-simple" in your submission area, and assembling the job submission scripts there. There will be no interaction with the PanDA server.

$ shrek --no-submit --no-check --no-handshake --tag sP00a-simple tests/simple-chain/*.yaml

There are three flags -- no-submit, no-check and no-handshake -- which instruct shrek not to have any interaction with the PanDA servers. On a typical call, shrek will begin by trying to send a handshake message to the PanDA server. PanDA will either respond that it is ready to accept jobs, or request authentication. (Or possibly fail at this point).

Run

$ shrek --no-submit --no-check --handshake --tag sP00a-simple tests/simple-chain/*.yaml

to make sure you are authenticated with the PanDA server, following any instructions you recieve. [You could omit the --handshake option as it is the default].

Next we want to validate the workflow. Run

$ shrek --no-submit --check --handshake --tag sP00a-simple tests/simple-chain/*.yaml

The PanDA server will return true if the CWL document is properly structured and able to be accepted. This is not a guarentee that the code will run, but it does demonstrate that shrek has converted the job definition files under tests/simple-chain into something which PanDA can handle. [Again, the --check flag is enabled by default and could be omitted.]

If all goes well, shrek should report that PanDA considers the workflow to be valid, and provides instructions for how to submit the job by hand.

Finally let's submit the job to PanDA via shrek. Issue

$ shrek --submit --tag sP00a-simple tests/simple-chain/*.yaml

As mentioned above, the handshake and check options are on by default, so we have omitted them from this example. The submit option, however, is not on by default. So we explicitly tell shrek to submit the job.

If all goes well, you'll see a URL pointing you to the job monitoring page for your run.

Troubleshooting

What to do if something goes wrong

Python complains about missing executables for graphviz

If you see a message FileNotFoundError: [Errno 2] No such file or directory: 'dot', the graphviz executable is not installed on the system. This is used to create a diagram of the workflow chain, and is not required for the construction of the files needed for PanDA submission. You can add the option

--no-diagram

to the command line. This will supress generation of the workflow diagram and allow submission to proceed.

n.b. --no-diagram is now the default option

PanDA did not validate the workflow. Submission canceled.

This message is returned when PanDA recieives and invalid workflow document, but will also occur if there is a problem with authenticating to either or both the PanDA and rucio servers.

SHREK assumes that your PanDA and rucio account has the same name as your account on the computer that you submit from. If this is not the case, you can provide the username as

--user my-username

The submission goes through, but PanDA workflow monitor reports "doSetup failed with local variable 'child_process' referenced before assignment failed to setup task"

I see this error somewhat frequently. Give it some time and PanDA may recover.