Offset patcher

[olessia]

As a review, could you please checkout the branch and follow the process below. This should uncover any issues/things to improve.

I'll also document this in our playbook.

Fixing stuck offsets

• Unschedule import
• Download the offsets, e.g. HADOOP_CONF_DIR=<path_to_starscream_repo>/.dev/starscream/spark/current/conf/conf.cloudera.yarn/ hadoop fs -copyToLocal /var/camus/execution/history/2018-01-01-00-00-00 .
• Running the script, e.g java -Xms1G -Xmx2G -cp/usr/local/Cellar/hadoop/2.8.0/bin/hadoop classpath:/Users/olessia/src/github.com/Shopify/camus/camus-shopify/target/camus-shopify-0.1.0-shopify1.jar org.wikimedia.analytics.refinery.job.CamusOffsetPatcher -e /Users/olessia/Documents/kafka_offsets/2018-02-20-13-30-30 -t nginx.shopify -p 120 -o 1226657431
• Upload new offset file to management2
• Sanity check of old offset file vs new offset file e.g. hadoop fs -libjars /u/apps/camus/current/camus-shopify-0.1.0-shopify1.jar -text file:///<path>/patched-offsets-m-00104
• Delete old file, replace with new file with the same name
• Start camus, check offsets

[dterror-zz]

(we can go straight to prod, too, I think)