Example HTTPFS communication using curl+kerberos

gistfile1.sh

#!/bin/bash

#
# Uses a keytab to auth to kerberos and then creates an empty file/directory
# to notify oozie that data is ready.
# Specify the name of the dataset to notify for as the first argument,
# and optionally the day in UTC during which the data is intended to be loaded
# (defaults to the current day in UTC)
# e.g. :
#
# /path/to/datasci_notify.sh audopt_us
# OR
# /path/to/datasci_notify.sh audopt_us 2013-03-23
#

KEYTAB=`dirname $0`/datasci.keytab
PRINCIPAL=datasci@COLLECTIVE-MEDIA.NET
KINIT=/usr/bin/kinit
CURL=/usr/bin/curl
NOTIFY_PATH=/datasciences/batchimport
READY_FILE=_READY
HTTPFS="http://nn01-301.ny7.collective-media.net:14000/webhdfs/v1"
function die {
  echo "ERROR: $1"
  exit 1 
}

DATASET_NAME=$1
[ -z "$DATASET_NAME" ] && die "Please specify name of dataset to notify for as the first argument!"

UTC_DAY="$2"
if [ -z "$UTC_DAY" ]; then
  TZ=UTC
  UTC_DAY=`date +'%Y-%m-%d'`
  echo "Using default date of $UTC_DAY UTC"
fi

[ -f "$KEYTAB" ] || die "keytab file $KEYTAB does not exist!"
[ -x "$KINIT" ] || die "kinit binary $KINIT does not exist or is not executable!"
[ -x "$CURL" ] || die "curl command $CURL does not exist or is not executable!"
 
$KINIT -k -t "$KEYTAB" "$PRINCIPAL" || die "Failed to authenticate to kerberos using keytab $KEYTAB as $PRINCIPAL"

TARGET_DIR="${NOTIFY_PATH}/${DATASET_NAME}/${UTC_DAY}"
$CURL -f -X PUT -i --negotiate -u : "${HTTPFS}${TARGET_DIR}?op=MKDIRS&permission=1775" || die "Failed to create target dir ${TARGET_DIR} in HDFS"
$CURL -f -X PUT -i --negotiate -u : "${HTTPFS}${TARGET_DIR}/${READY_FILE}?op=CREATE&overwrite=true" && \
  $CURL -f -H "Content-Type: application/octet-stream" -X PUT -i --negotiate -u : "${HTTPFS}${TARGET_DIR}/${READY_FILE}?op=CREATE&overwrite=true&data=true" || die "Failed to create ${TARGET_DIR}/${READY_FILE}" || \
    die "Failed to create ${TARGET_DIR}/${READY_FILE}"

echo "${TARGET_DIR}/${READY_FILE} created"

Thanks. We need to transfer the file with .tmp extention while the file is in transit and will have to remove the .tmp extention once transfer is complete. Can you please suggest how can we achive it using curl