surgical script that kicks off user defined compactions to purge tombstones in c*

user-defined-compaction.sh

#!/bin/bash

#set -x

DATA_DIRECTORY="/var/lib/cassandra/data"

CASS_TOOLS_BIN_DIRECTORY="/DSE_DIR_HERE/resources/cassandra/tools/bin/"

KEYSPACE_NAME="assethub"
TABLE_NAME="clusters"

SSTABLE_DIR="$DATA_DIRECTORY/$KEYSPACE_NAME/$TABLE_NAME"

MIN_FILE_SIZE_KB=100
TOMBSTONE_THRESHOLD=.05

#DEBUG=true

cd /tmp || exit

if [ ! -e jmxsh-R5.jar ]; then
  wget https://storage.googleapis.com/google-code-archive-downloads/v2/code.google.com/jmxsh/jmxsh-R5.jar
fi


# Names must not have spaces or globable chars
SSTABLE_ARRAY=($(find "$SSTABLE_DIR"* -maxdepth 1 -size +"$MIN_FILE_SIZE_KB"k -name '*Data.db'))

TOMBSTONE_ARRAY=($(find "$SSTABLE_DIR"* -maxdepth 1 -size +"$MIN_FILE_SIZE_KB"k -name '*Data.db' | xargs "$CASS_TOOLS_BIN_DIRECTORY"sstablemetadata 2> /dev/null | grep tombstones: | awk -F':' '{print $2}'))

#DEBUG
if [ -n "$DEBUG" ]; then
  echo "SSTable array ${SSTABLE_ARRAY[*]}"
  echo "Tombstone array ${TOMBSTONE_ARRAY[*]}"
fi

for i in "${!TOMBSTONE_ARRAY[@]}"
do
  if [ -n "$DEBUG" ]; then
    echo "Tombstone count, Threshold: ${TOMBSTONE_ARRAY[$i]}, $TOMBSTONE_THRESHOLD"
  fi

  if (( $(echo "${TOMBSTONE_ARRAY[$i]} $TOMBSTONE_THRESHOLD" | awk '{print ($1 > $2)}') )); then
    echo "jmx_invoke -m org.apache.cassandra.db:type=CompactionManager forceUserDefinedCompaction " "${SSTABLE_ARRAY[$i]}" > UDC.sh
  fi
done

java -jar jmxsh-R5.jar -h localhost -p 7199 -q UDC.sh