I want a shell scripts for find which kubernetes pods is newly added to the cluster. Include the pod name,date&Time and Namespace.
I have tried with the following bash script:
#!/bin/bash
p=50 #total pods count
pcount=`kubectl get pods |wc -l`
###Addition check
if [ $ncount -gt $n ]
then
####pods variable have all pods name.
pods=`kubectl get pods|awk '{print $1}'|awk '{if(NR>1)print}'| xargs`
###start variable have all dates pods created
start=`kubectl describe pod $pods |grep "Start Time"|awk '{print $3 $4 $5 $6}'`
###max variable have total number of pods
max=`kubectl get pods |awk '{if(NR>1)print}'| wc -l`
dt=`date +%a,%d%b%Y`
array=( $start )
i=0
while [ $i -lt $max ];
do
# echo "inside loop ${array[$i]}"
if [[ "$dt" == "${array[$i]}" ]];then
dat=`date "+%a, %d %b %Y"`
name=`kubectl describe pod $pods |grep -v SecretName: |echo "$dat" |egrep 'Name|Start Time'`
printf "\n"
echo "Newly Added pods are: $name"
fi
i=$(( $i + 1 ))
done
fi
The script working almost fine. But I need only today's created pods,the scripts showing all pods Name, Start Time and Namespace.
Please help.
Your script has numerous issues and inefficiencies. Repeatedly calling a somewhat heavy command like kubectl should be avoided; try to rearrange things so that you only run it once, and extract the information you need from it. I'm vaguely guessing you actually want something along the lines of
#!/bin/bash
# Store pod names in an array
pods=($(kubectl get pods |
awk 'NR>1 { printf sep $1; sep=" "}'))
if [ ${#pods[@]} -gt $n ]; then # $n is still undefined!
for pod in "${pods[@]}"; do
kubectl describe pod "$pod" |
awk -v dt="$(date +"%a, %d %b %Y")" '
/SecretName:/ { next }
/Name:/ { name=$NF }
/Start Time:/ { t=$3 $4 $5 $6;
if (t==dt) print name
name="" }'
done
fiOnce you run Awk anyway, it makes sense to refactor as much as the processing into Awk; it can do everything grep and cut and sed can do, and much more. Notice also how we use the $(command) command substitution syntax in preference over the obsolescent legacy `command` syntax.
kubectl with -o=json would probably be a lot easier and more straightforward to process programmatically so you should really look into that. I don't have a Kubernetes cluster to play around with so I'm only pointing this out as a direction for further improvement.