Run MongoDB on Arm-based Azure Cobalt 200 instances
Introduction
What are Cobalt 200 and MongoDB?
Create an Arm-based cloud virtual machine using Cobalt 200
Install MongoDB and Mongosh
MongoDB baseline testing
Generate a MongoDB workload
Monitor MongoDB with mongotop
Monitor MongoDB with mongostat
Next Steps
Run MongoDB on Arm-based Azure Cobalt 200 instances
Install MongoDB Database Tools
The tools package provides mongotop and mongostat. Download and install the Ubuntu 24.04 Arm64 package:
wget https://fastdl.mongodb.org/tools/db/mongodb-database-tools-ubuntu2404-arm64-100.13.0.deb
sudo apt install -y ./mongodb-database-tools-ubuntu2404-arm64-100.13.0.deb
mongotop --version
mongostat --version
The .deb installs the tools into the system PATH; no shell profile changes are needed.
Create a bounded workload
In terminal 1 on your VM, create long_system_load.js with this content. The script uses only a tutorial collection, seeds 1,000 documents, and runs insert, read, update, and delete operations for 60 seconds:
const workloadDB = db.getSiblingDB("benchmarkDB");
const collection = workloadDB.testCollection;
for (let i = 0; i < 1000; i++) {
collection.updateOne({ _id: i }, { $set: { value: i } }, { upsert: true });
}
const deadline = Date.now() + 60000;
let cycles = 0;
while (Date.now() < deadline) {
const id = cycles % 1000;
const inserted = collection.insertOne({ value: cycles, temporary: true });
assert(inserted.acknowledged);
assert(collection.findOne({ _id: id }) !== null);
assert(collection.updateOne({ _id: id }, { $inc: { value: 1 } }).matchedCount === 1);
assert(collection.deleteOne({ _id: inserted.insertedId }).deletedCount === 1);
cycles++;
}
print("Completed workload cycles:", cycles);
Run the script from your VM terminal:
mongosh --quiet mongodb://127.0.0.1:27017 long_system_load.js
The script stops after its 60-second loop and reports the number of completed cycles. Keep it running while you sample the monitoring tools in terminals 2 and 3. Connect each terminal to the same VM using SSH. Re-run the script if it finishes before you start monitoring.
This is a single-client synthetic workload for learning the monitoring tools. A cycle contains four operations; cycle count is not a MongoDB capacity measurement. The script generates activity in benchmarkDB.testCollection, not internal replication or transaction-system collections.