Skip to content

Latest commit

 

History

History
212 lines (149 loc) · 9.31 KB

README.md

File metadata and controls

212 lines (149 loc) · 9.31 KB

TODO

  • add pick events support

Installation

  • Clone this repository into your ~/public_html/ directory.
  • Get a valid proxy. I recommend making this auto-renewing (see section below).
  • Edit usernames appropriately in config.py. And make a passwordless key as indicated in the comments.
  • Edit the port in serve.py if needed and then run ./start.sh to start the flask backend
  • Depending on which uaf and port the backend is running on, edit the BASEURL in dis_client.py and js/main.js.

Crontab/auto-renewing proxy

  • Follow instructions from this Twiki to create a handy auto-renewing proxy.

  • As dis uses a user proxy to make queries, you need to make sure there's always a fresh one at your disposal. Put this into your crontab (crontab -e)

* 55 15 * * * /home/users/${USER}/cron/renewProxy.sh >/dev/null 2>&1

This should correspond to the long-lived proxy setup.

Testing

Unit tests can be run in the same environment used in start.sh:

source /cvmfs/cms.cern.ch/cmsset_default.sh
cd /cvmfs/cms.cern.ch/slc6_amd64_gcc630/cms/cmssw/CMSSW_9_4_9; 
cmsenv; 
cd -; 
python tests.py
python dis_client.py -e -

This will run unit tests on the backend (tests.py) and on the client (dis_client.py).

Usage instructions and examples

A query has 3 parts: the query string, the query type, and the boolean "short" option (which governs the level of detail that is returned by the API).

General notes

  • Wildcards are accepted (*) for most queries where you'd want to use wildcards. Just try it.
  • dis_client.py syntax will be used here, but of course they have obvious mappings to use them on the website
  • I recommend putting dis_client.py on your PATH (and PYTHONPATH) somewhere.

Query types

basic

dis_client.py -t basic /GJets_HT-600ToInf_TuneCUETP8M1_13TeV-madgraphMLM-pythia8/RunIISpring15DR74-Asympt25ns_MCRUN2_74_V9-v1/MINIAODSIM

  • Here, you will see an output of the number of events in the dataset, the number of files, number of lumi blocks, and dataset file size.
  • Also note that the -t basic is default, so you don't need to do it for this basic query type.

dis_client.py "/GJets_HT-*_TuneCUETP8M1_13TeV-madgraphMLM-pythia8/RunIISpring15DR74-Asympt25ns_MCRUN2_74_V9-v1/MINIAODSIM"

  • The wildcard now will cause the output to be a list of matching dataset names.
  • If you put ,all after the dataset pattern, you will also see the samples not yet marked as "valid" in DBS

dis_client.py --detail "/GJets_HT-*_TuneCUETP8M1_13TeV-madgraphMLM-pythia8/RunIISpring15DR74-Asympt25ns_MCRUN2_74_V9-v1/MINIAODSIM"

  • Specifying --detail (equivalently, unchecking short in the web interface) will show the number of events, number of files, etc. for each dataset matching the wildcard.

files

dis_client.py -t files /GJets_HT-600ToInf_TuneCUETP8M1_13TeV-madgraphMLM-pythia8/RunIISpring15DR74-Asympt25ns_MCRUN2_74_V9-v1/MINIAODSIM

  • This will show a list of a handful of the files from this dataset, along with filesize and number of events. To show all the files, provide the --detail option (note that sometimes, you can get thousands of files since datasets can be large, which is why the default is only a handful, which suits my main use case which is to check one file in a dataset for something specific).

config

dis_client.py -t files /GJets_HT-600ToInf_TuneCUETP8M1_13TeV-madgraphMLM-pythia8/RunIISpring15DR74-Asympt25ns_MCRUN2_74_V9-v1/MINIAODSIM

  • Shows information about the CMSSW version and global tag used for processing this dataset

mcm

dis_client.py -t mcm /GJets_HT-600ToInf_TuneCUETP8M1_13TeV-madgraphMLM-pythia8/RunIISpring15DR74-Asympt25ns_MCRUN2_74_V9-v1/MINIAODSIM

  • This shows basic MCM information (for ALL information, throw it the detail option) for the OLDEST PARENT of the dataset (presumably GENSIM)
  • This includes the fragment, cross-section, matching/filter efficiency, CMSSW release, MCDB ID, and dataset status (done, running, etc.)

dis_client.py -t mcm "/GJets_HT-600ToInf_TuneCUETP8M1_13TeV-madgraphMLM-pythia8/RunIISpring15DR74-Asympt25ns_MCRUN2_74_V9-v1/MINIAODSIM,this"

  • To get MCM information for the actual dataset you feed in, tack on the ,this modifier

chain

dis_client.py -t chain /GJets_HT-600ToInf_TuneCUETP8M1_13TeV-madgraphMLM-pythia8/RunIISpring15DR74-Asympt25ns_MCRUN2_74_V9-v1/MINIAODSIM

  • This recurses through the chained request for the sample and prints out all driver commands/fragments. It's useful if you want to go from GENSIM to MINIAODSIM without clicking 30 times in the MCM interface to get the necessary driver commands.

parents

dis_client.py -t parents /GJets_HT-600ToInf_TuneCUETP8M1_13TeV-madgraphMLM-pythia8/RunIISpring15DR74-Asympt25ns_MCRUN2_74_V9-v1/MINIAODSIM

  • Returns a list of datasets found as we recurse up the tree of parenthood

snt

dis_client.py -t snt "/GJets_HT-*_TuneCUETP8M1_13TeV-madgraphMLM-pythia8/RunIISpring15DR74-Asympt25ns_MCRUN2_74_V9-v1/MINIAODSIM"

  • Much like the other database queries, this just uses the SNT datasets (returns information about ntupled nevents, cross-section, kfactor, hadoop location, etc.)
  • The --detail option just provides more details like the filter type, twiki name, who the sample was assigned to, etc.

runs

dis_client.py -t runs "/SinglePhoton/Run2016E-PromptReco-v2/MINIAOD"

  • Returns a list of runs contained within the dataset
  • Hint: try dis_client.py -t runs "/SinglePhoton/Run2016E-PromptReco-v2/MINIAOD | stats" to see the first, last, and number of runs.

sites

dis_client.py -t sites "/MET/Run2016D-PromptReco-v2/MINIAOD"

  • Gives phedex information about a dataset or even a /store/ file, including the replication sites.

Selectors/modifiers, greppers, and all that

dis_client.py -t snt "/gjet*,cms3tag=*V08*,gtag=*v3 | grep location,cms3tag,gtag,xsec"

  • This uses a selector and grepper to show all Gjet SNT samples that have a cms3tag containing V08 and global tag ending with v3. The "cms3tag" name comes from the return value when doing a normal query, so you can find all the values that are selectable by making an inclusive query. Same goes for the grep fields which limit what is shown to you in the return information.
  • But Nick, this looks ugly and I can't copy and paste the output easily into other scripts. It would be nice if we could put the same information for each sample on the same line. Fear not. There is a --table option which puts the results into a pretty table with nice columns.

dis_client.py -t snt "/gjet*,cms3tag=*V08*,gtag=*v3 | grep location,cms3tag,gtag,xsec" --table

For a more practical application, what if we want to get the total number of events for all HT binned GJet samples in CMS3?

dis_client.py -t snt "/gjet*ht-*/*/*,cms3tag=*V08* | grep nevents_out"

  • This is a start. We see one column of numbers with the "nevents_out" field. What if we could add them together and display some statistics?

dis_client.py -t snt "/gjet*ht-*/*/*,cms3tag=*V08* | grep nevents_out | stats"

  • Piping it into stats, we get the number of entries, the total, the minimum, and maximum for a list of numbers. More generally, any list of numbers can be piped into stats. Same with the run query type above (if we wanted to find the first or last run in a dataset, for example).

  • There is also a sort.

SQL tips

  • Dump the SNT .db file to a .csv for quick grepping/visual inspection with
# dump .db to .csv
python -c 'from snt_db import SNTDBInterface; db = SNTDBInterface(fname="allsamples.db"); db.export_to_csv("temp.csv"); db.close()'
# and then you can edit `temp.csv` or do whatever in plaintext
# load .csv back into .db
python -c 'from snt_db import SNTDBInterface; db = SNTDBInterface(fname="allsamples.db"); db.load_from_csv("temp.csv"); db.close()'
  • (for pure bash, sqlite3 -header -csv allsamples.db "select * from sample;" > out.csv)
  • There are many examples of usage at the bottom of db.py, in db_tester.py and in the snt/update_snt/delete_snt sections of api.py.

Misc operational tips

  • If you want to move a published dataset to another folder, you should make DIS aware of this. The simplest query to update a location would be
import dis_client as dc
dsname="/ZZ_TuneCUETP8M1_13TeV-pythia8/RunIISummer16MiniAODv2-PUMoriond17_80X_mcRun2_asymptotic_2016_TrancheIV_v6-v1/MINIAODSIM"
cms3tag="CMS3_V08-00-16"
location="/hadoop/cms/store/group/snt/my/fake/dir/"
print(dc.query(
    "sample_type=CMS3,dataset_name={dsname},cms3tag={cms3tag},location={location}".format(dsname=dsname,cms3tag=cms3tag,location=location),
    typ="update_snt"
)["payload"])

since you need a sample_type, dataset_name, and cms3tag to uniquely specify a sample, and then the rest is used to update the existing entry. Of course for new entries, you should specify a lot more.

API Usage

The primary purpose of this was to provide programmatic access to DBS, MCM, DAS, etc, so the --json option can be passed to any dis_client.py query to output a json. Even better, if it's on your path, you can import it directly from python and get a dictionary back as the response

import dis_client
response =  dis_client.query(q="..." [, typ="basic"] [, detail=False])
data = response["payload"]
print response