kubeshop
Support Document Retention Policy Configuration for MongoDB via Helm Chart
Description
We recently encountered performance issues with our Testkube deployment, which we traced back to the MongoDB instance used for storing workflow execution results. The root cause was the unbounded growth of the testworkflowresults collection:
> db.testworkflowresults.countDocuments()
// 218361
Following the Testkube MongoDB administration documentation, we manually implemented a TTL index to purge old execution data:
db.testworkflowresults.createIndex(
{ createdAt: 1 },
{ expireAfterSeconds: 604800 } // 7 days
)
After this change, the document count dropped and performance significantly improved:
> db.testworkflowresults.countDocuments()
// 10671
To automate this for future installations, we created an initdbScript to apply this TTL index. This introduces extra complexity and maintenance burden and we are not really sure if this is the correct or only this available automation approach
mongodb:
initdbScripts:
create-indexes.sh: |
#!/bin/bash
until mongosh --eval "print(\"waited for connection\")" > /dev/null 2>&1; do
sleep 2
done
TIMEOUT=300
START_TIME=$(date +%s)
TIMEOUT_REACHED=false
until mongosh --eval '
db = db.getSiblingDB("testkube");
if (db.getCollectionNames().length > 0) {
print("testkube database is ready");
exit(0);
} else {
print("waiting for testkube database...");
exit(1);
}
' > /dev/null 2>&1; do
CURRENT_TIME=$(date +%s)
ELAPSED_TIME=$((CURRENT_TIME - START_TIME))
if [ $ELAPSED_TIME -ge $TIMEOUT ]; then
echo "Timeout reached: testkube database not available after 5 minutes"
TIMEOUT_REACHED=true
break
fi
echo "Waiting for testkube database to be available... (${ELAPSED_TIME}s elapsed)"
sleep 5
done
if [ "$TIMEOUT_REACHED" = true ]; then
echo "Exiting due to timeout"
exit 1
fi
mongosh --eval '
db = db.getSiblingDB("testkube");
db.testworkflowresults.createIndex(
{ createdAt: 1 },
{ expireAfterSeconds: 604800 }
);
'
Feature Request:
Instead of requiring manual TTL index creation or custom init scripts, we propose that Testkube should support a retention policy configuration for data stored in MongoDB, defined via Helm chart values. Internally, Testkube could then translate this into the appropriate TTL index creation logic.
testkube:
dataRetention:
enabled: true
duration: 7d # Automatically remove results older than 7 days
This would allow users to define their retention strategy at the configuration level without needing to understand the underlying MongoDB index mechanics.
Thank you for the excellent work on Testkube—we’d love to see this improvement considered!
Thank you @JaroVojtek Great suggestiion! For @olensmar and @jmorante-ks to prioiritise
Hi @JaroVojtek - thanks for this request; it's in our backlog, making its way through the process.. very sorry to keep you waiting 🙏
