diff options
| -rw-r--r-- | README.md | 37 | ||||
| -rw-r--r-- | python/README.md | 52 |
2 files changed, 46 insertions, 43 deletions
@@ -6,34 +6,29 @@ \ooooooo| |___/\__,_|_| |_|\__,_|\___|_| \__,_| \_/\_/ |_|\___|_| -This repo contains hadoop tasks (mapreduce and pig), luigi jobs, and other -scripts and code for the internet archive (web group) journal ingest pipeline. +This repo contains hadoop jobs, luigi tasks, and other scripts and code for the +internet archive web group's journal ingest pipeline. -This repository is potentially public. +Code in tihs repository is potentially public! Archive-specific deployment/production guides and ansible scripts at: [journal-infra](https://git.archive.org/bnewbold/journal-infra) -## Python Setup +**./python/** contains Hadoop streaming jobs written in python using the +`mrjob` library. Most notably, the **extraction** scripts, which fetch PDF +files from wayback/petabox, process them with GROBID, and store the result in +HBase. -Pretty much everything here uses python/pipenv. To setup your environment for -this, and python in general: +**./scalding/** contains Hadoop jobs written in Scala using the Scalding +framework. The intent is to write new non-trivial Hadoop jobs in Scala, which +brings type safety and compiled performance. - # libjpeg-dev is for some wayback/pillow stuff - sudo apt install -y python3-dev python3-pip python3-wheel libjpeg-dev build-essential - pip3 install --user pipenv +**./pig/** contains a handful of Pig scripts, as well as some unittests +implemented in python. -On macOS: +## Running Hadoop Jobs - brew install libjpeg pipenv +The `./please` python3 wrapper script is a helper for running jobs (python or +scalding) on the IA Hadoop cluster. You'll need to run the setup/dependency +tasks first; see README files in subdirectories. -Each directory has it's own environment. Do something like: - - cd python - pipenv install --dev - pipenv shell - -## Possible Issues with Setup - -Bryan had `~/.local/bin` in his `$PATH`, and that seemed to make everything -work. diff --git a/python/README.md b/python/README.md index aebc160..5c83908 100644 --- a/python/README.md +++ b/python/README.md @@ -3,9 +3,20 @@ Hadoop streaming map/reduce jobs written in python using the mrjob library. ## Development and Testing -System dependencies in addition to `../README.md`: +System dependencies on Linux (ubuntu/debian): -- `libjpeg-dev` (for wayback libraries) + sudo apt install -y python3-dev python3-pip python3-wheel libjpeg-dev build-essential + pip3 install --user pipenv + +On macOS (using Homebrew): + + brew install libjpeg pipenv + +You probably need `~/.local/bin` on your `$PATH`. + +Fetch all python dependencies with: + + pipenv install --dev Run the tests with: @@ -16,10 +27,19 @@ Check test coverage with: pytest --cov --cov-report html # open ./htmlcov/index.html in a browser -TODO: Persistant GROBID and HBase during development? Or just use live -resources? +## Running Python Jobs on Hadoop + +The `../please` script automates these steps; you should use that instead. -## Extraction Task +When running python streaming jobs on the actual hadoop cluster, we need to +bundle along our python dependencies in a virtual env tarball. Building this +tarball can be done like: + + export PIPENV_VENV_IN_PROJECT=1 + pipenv install --deploy + tar -czf venv-current.tar.gz -C .venv .""" + +### Extraction Task An example actually connecting to HBase from a local machine, with thrift running on a devbox and GROBID running on a dedicated machine: @@ -30,13 +50,7 @@ running on a devbox and GROBID running on a dedicated machine: --grobid-uri http://wbgrp-svc096.us.archive.org:8070 \ tests/files/example.cdx -Running from the cluster: - - # Create tarball of virtualenv - export PIPENV_VENV_IN_PROJECT=1 - pipenv shell - export VENVSHORT=`basename $VIRTUAL_ENV` - tar -czf $VENVSHORT.tar.gz -C /home/bnewbold/.local/share/virtualenvs/$VENVSHORT . +Running from the cluster (once a ./venv-current.tar.gz tarball exists): ./extraction_cdx_grobid.py \ --hbase-table wbgrp-journal-extract-0-qa \ @@ -44,10 +58,10 @@ Running from the cluster: --grobid-uri http://wbgrp-svc096.us.archive.org:8070 \ -r hadoop \ -c mrjob.conf \ - --archive $VENVSHORT.tar.gz#venv \ + --archive venv-current.tar.gz#venv \ hdfs:///user/bnewbold/journal_crawl_cdx/citeseerx_crawl_2017.cdx -## Backfill Task +### Backfill Task An example actually connecting to HBase from a local machine, with thrift running on a devbox: @@ -57,18 +71,12 @@ running on a devbox: --hbase-host wbgrp-svc263.us.archive.org \ tests/files/example.cdx -Actual invocation to run on Hadoop cluster (running on an IA devbox, where -hadoop environment is configured): - - # Create tarball of virtualenv - export PIPENV_VENV_IN_PROJECT=1 - pipenv install --deploy - tar -czf venv-current.tar.gz -C .venv . +Running from the cluster (once a ./venv-current.tar.gz tarball exists): ./backfill_hbase_from_cdx.py \ --hbase-host wbgrp-svc263.us.archive.org \ --hbase-table wbgrp-journal-extract-0-qa \ -r hadoop \ -c mrjob.conf \ - --archive $VENVSHORT.tar.gz#venv \ + --archive venv-current.tar.gz#venv \ hdfs:///user/bnewbold/journal_crawl_cdx/citeseerx_crawl_2017.cdx |