pcluster-manager
pcluster-manager copied to clipboard
Cost-tags.sh script does not treat aws tagging API throttling
I noticed that when we launch large jobs (like 100 nodes) not all instances are getting tagged by the tagging script perhaps because of the throttling on the AWS API endpoint.
I suggest spreading the tagging for a short interval instead (maybe randomize within 2 minutes timeframe) such as
...
if [ ${update} -eq 1 ]; then
sleep $(( $RANDOM % 120 + 1 ))
# Instance ID
...