Get AWS RDS Enhanced Monitoring statistics from CloudWatch and show something similar to Linux top command

rds-top.sh

#!/bin/bash
# rds-top.sh
#   by <[email protected]>, 20220616
#     * Changed some --start-time syntax
#     * Added --profile and --region
#     * Made nicer numbers for memory/swap
#     * Fixed disk IO output to only show devices with real data
#   by <[email protected]>, 20190822
#   based on the work of Matheus de Oliveira <[email protected]>
#
# WISHLIST
# 
# As this floats around the Intertubes, tack on some features you'd like
# others to think about and add:
# * ncurses / simplecurses like interface to make it more top-like 
# * start at a time in the past and replay through the seconds 
#
set -e

function usage()
{
    cat << EOF
Usage: $0 [options] rds_instance_id

OPTIONS:
    -s or --start-time=t           Optional: Specify the start time in seconds since the Unix epoch
                                     OR specify timestamp parsable by your GNU date utility
    -m or --sort-by-mem            Optional: Sorts output by memory. Default is to sort by CPU
    --profile=me                   AWS CLI Profile
    --region=us-east...            AWS CLI Region
    
EXAMPLES:

  rds_top my-prod-writer
  rds_top --start-time "13 days ago" my-prod-readonly
  rds_top --start-time 1655318348 my-prod-readonly
  rds_top --profile me@account --region us-west-2 --sort-by-mem failover-instance
  # a more top-like real time view
  watch --interval=1 "rds_top my-instance"

EOF
}


if [ $# -eq 0 ]
then
    usage
    exit
fi


for i in "$@"
do
case $i in
    # removing = token requirement for looser arguments
    #-s=*|--start-time=*)
    #    START_TIME="${i#*=}"
    -s | --start-time )
        START_TIME="$2"
        shift 2
        ;;
    -m | --sort-by-mem )
        SORT_BY_MEM=1
        shift
        ;;
    --profile )
        PROFILE="--profile $2"
        shift 2
        ;;
    --region )
        REGION="--region $2"
        shift 2
        ;;
    *)
        # unknown option
    ;;
esac
done
INSTANCE_ID="$1"
if [[ "$INSTANCE_ID" == "" ]]
then
    echo "You have to supply an instance ID!"
    usage
    exit
fi

if [[ "$START_TIME" != "" && ! "$START_TIME" =~ ^[[:digit:]]+$ ]]; then
  START_TIME=$(date --date="$START_TIME" '+%s')
fi

AWS="aws $PROFILE $REGION"

# Get resource id
#echo "$AWS rds describe-db-instances --db-instance-identifier "$INSTANCE_ID" --output json | jq -r '.DBInstances[].DbiResourceId'"
resource_id="$($AWS rds describe-db-instances --db-instance-identifier "$INSTANCE_ID" --output json | jq -r '.DBInstances[].DbiResourceId')"
[[ "$resource_id" == "" ]] && exit 1

# Build logs parameters
params=(--log-group-name RDSOSMetrics --log-stream-name "$resource_id" --limit 1 --output json)
if [[ "$START_TIME" =~ ^[[:digit:]]+$ ]];
then
  params+=(--start-time "$((START_TIME * 1000))" --start-from-head)
fi

# Get log events
message_json=$($AWS logs get-log-events "${params[@]}" | jq -r '.events[].message')
#echo "$message_json" | jq '.'

echo "${message_json}" |
  jq -r '"\(.instanceID) - \(.timestamp) - \(.uptime) up, load average: \(.loadAverageMinute.one), \(.loadAverageMinute.five), \(.loadAverageMinute.fifteen)
Tasks: \(.tasks.total) total, \(.tasks.running) running, \(.tasks.sleeping) sleeping, \(.tasks.stopped) stopped, \(.tasks.zombie) zombie
%Cpu(s): \(.cpuUtilization.user) us, \(.cpuUtilization.system) sy, \(.cpuUtilization.nice) ni, \(.cpuUtilization.idle) id, \(.cpuUtilization.wait) wa, \(.cpuUtilization.steal) st
MiB Mem: \(.memory.total / 1024|floor) total, \(.memory.free / 1024|floor) free, \((.memory.total - .memory.free) / 1024|floor) used, \((.memory.cached + .memory.buffers) / 1024|floor) buff/cache
MiB Swap: \(.swap.total / 1024|floor) total, \(.swap.free / 1024|floor) free, \(.swap.cached|floor) cached"'

echo

echo "${message_json}" |
  jq -r '.network[] | "Net \(.interface): \(.rx) rx, \(.tx) tx"'

# Mimic `iostat -x`
echo "${message_json}" |
  jq -r '.diskIO[] | select('.device') | "Disk \(.device): \(.tps) tps, \(.rrqmPS) rrqm/s, \(.wrqmPS) wrqm/s, \(.writeKbPS) wKB/S, \(.readKbPS) rKB/S, \(.avgReqSz) avgrq-sz, \(.avgQueueLen) avgqu-sz, \(.await) await, \(.util) %util"'

echo

sort_by=".cpuUsedPc"
if [[ $SORT_BY_MEM -eq 1 ]]
then
  sort_by=".memoryUsedPc"
fi

echo "${message_json}" |
  jq -r '["PID", "PPID", "VSS", "RSS", "%CPU", "%MEM", "COMMAND"], (.processList | sort_by('"$sort_by"') | reverse[] | [.id, .parentID, .vss, .rss, .cpuUsedPc, .memoryUsedPc, .name]) | @tsv' |
  column -t -s $'\t'