diff --git a/lib/dstat b/lib/dstat
index 4b22752ced..f11bfa55c0 100644
--- a/lib/dstat
+++ b/lib/dstat
@@ -21,11 +21,17 @@ function start_dstat {
     # A better kind of sysstat, with the top process per time slice
     DSTAT_OPTS="-tcmndrylpg --top-cpu-adv --top-io-adv"
     run_process dstat "dstat $DSTAT_OPTS"
+
+    # To enable peakmem_tracker add:
+    #    enable_service peakmem_tracker
+    # to your localrc
+    run_process peakmem_tracker "$TOP_DIR/tools/peakmem_tracker.sh"
 }
 
 # stop_dstat() stop dstat process
 function stop_dstat {
     stop_process dstat
+    stop_process peakmem_tracker
 }
 
 # Restore xtrace
diff --git a/tools/peakmem_tracker.sh b/tools/peakmem_tracker.sh
new file mode 100755
index 0000000000..0d5728a538
--- /dev/null
+++ b/tools/peakmem_tracker.sh
@@ -0,0 +1,96 @@
+#!/bin/bash
+#
+# Licensed under the Apache License, Version 2.0 (the "License"); you may
+# not use this file except in compliance with the License. You may obtain
+# a copy of the License at
+#
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
+# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
+# License for the specific language governing permissions and limitations
+# under the License.
+
+set -o errexit
+
+# time to sleep between checks
+SLEEP_TIME=20
+
+# MemAvailable is the best estimation and has built-in heuristics
+# around reclaimable memory.  However, it is not available until 3.14
+# kernel (i.e. Ubuntu LTS Trusty misses it).  In that case, we fall
+# back to free+buffers+cache as the available memory.
+USE_MEM_AVAILBLE=0
+if grep -q '^MemAvailable:' /proc/meminfo; then
+    USE_MEM_AVAILABLE=1
+fi
+
+function get_mem_available {
+    if [[ $USE_MEM_AVAILABLE -eq 1 ]]; then
+        awk '/^MemAvailable:/ {print $2}' /proc/meminfo
+    else
+        awk '/^MemFree:/ {free=$2}
+            /^Buffers:/ {buffers=$2}
+            /^Cached:/  {cached=$2}
+            END { print free+buffers+cached }' /proc/meminfo
+    fi
+}
+
+# whenever we see less memory available than last time, dump the
+# snapshot of current usage; i.e. checking the latest entry in the
+# file will give the peak-memory usage
+function tracker {
+    local low_point=$(get_mem_available)
+    while [ 1 ]; do
+
+        local mem_available=$(get_mem_available)
+
+        if [[ $mem_available -lt $low_point ]]; then
+            low_point=$mem_available
+            echo "[[["
+            date
+            echo "---"
+            # always available greppable output; given difference in
+            # meminfo output as described above...
+            echo "peakmem_tracker low_point: $mem_available"
+            echo "---"
+            cat /proc/meminfo
+            echo "---"
+            # would hierarchial view be more useful (-H)?  output is
+            # not sorted by usage then, however, and the first
+            # question is "what's using up the memory"
+            #
+            # there are a lot of kernel threads, especially on a 8-cpu
+            # system.  do a best-effort removal to improve
+            # signal/noise ratio of output.
+            ps --sort=-pmem -eo pid:10,pmem:6,rss:15,ppid:10,cputime:10,nlwp:8,wchan:25,args:100 |
+                grep -v ']$'
+            echo "]]]"
+        fi
+
+        sleep $SLEEP_TIME
+    done
+}
+
+function usage {
+    echo "Usage: $0 [-x] [-s N]" 1>&2
+    exit 1
+}
+
+while getopts ":s:x" opt; do
+    case $opt in
+        s)
+            SLEEP_TIME=$OPTARG
+            ;;
+        x)
+            set -o xtrace
+            ;;
+        *)
+            usage
+            ;;
+    esac
+done
+shift $((OPTIND-1))
+
+tracker