Fix IOPS Issues
On this page
System and disk alert conditions can indicate IOPS issues.
Alert Conditions
You can configure the following alert conditions in the project-level alert settings page to trigger alerts.
The Disk read IOPS on Data Partition is alert condition is
raised if the average number of disk read operations per second exceeds
the specified threshold.
The Disk write IOPS on Data Partition is alert condition is
raised if the average number of disk write operations per second
exceeds the specified threshold.
System: CPU (User) % is measures the CPU usage of the
processes on the node, normalized by the number of CPUs. This value is
scaled to a range of 0-100%.
Common Triggers
A few common events may trigger these alerts:
A one-time event, such as an index build.
Fix the Immediate Problem
Consider a few possible actions to help resolve these alerts:
Use the Atlas Performance Advisor to view slow queries and suggested indexes.
Review Indexing Strategies for possible further indexing improvements.
Note
You may need to temporarily increase your cluster IOPS to create new indexes. To change a cluster's IOPS, go to the Cluster Configuration page and:
Cloud ProviderTierPossible Actions
AWS
M10,M20AWS
M30or largerGoogle Cloud
M10or largerAzureM10or largerAnalyze Query Performance to review how your queries are using your indexes.
Increase hardware resources, such as instance size and IOPS, in the Cluster Configuration Page.
Implement a Long-Term Solution
Disk IOPS Burst Credits for Atlas Clusters on Azure
Atlas clusters deployed to Azure may use credit-based bursting, but the disk will burst only if it has burst credits accumulated in its credit bucket. Azure also offers an on-demand bursting model, where the disk bursts whenever its needs exceed its current capacity.
See the Azure Disk Bursting documentation for more information about how bursting for Azure disks works.
Conserve Burst Credits
If you regularly exceed your configured IOPS threshold, you can avoid depleting your burst credits by increasing your configured IOPS with any of the following actions:
Increase Cluster IOPS to increase the cluster's IOPS threshold.
Increase Storage Capacity. The IOPS threshold increases as you increase storage capacity. Clusters configured to use at least 1 TB of storage have baseline IOPS performance that is equal to or greater than the maximum burst performance. These volumes do not deplete burst credit balances.
Upgrade Cluster Tier. Larger cluster tiers include higher IOPS thresholds.
Note
Cluster tiers
M140and larger are deployed with at least 1 TB of storage capacity by default. Clusters with 1 TB or more of storage capacity do not deplete burst credit balances.
Monitor Your Progress
These are a few possible methods to observe high disk I/O:
The disk IOPS use from the Disk IOPS graph in the cluster metrics exceeds the provisioned IOPS from the Atlas cluster configuration page.
The Normalized System CPU metric has a high IOWait curve. IOWait measures the percentage of time the CPU is idle and waiting for an I/O operation to complete. The Normalized System CPU chart is located under the Hardware Metrics section of the Metrics tab.