Performance recommendations

Article
02/29/2024

The performance recommendations in Azure Advisor can help improve the speed and responsiveness of your business-critical applications. You can get performance recommendations from Advisor on the Performance tab of the Advisor dashboard.

Sign in to the Azure portal.
Search for and select Advisor from any page.
On the Advisor dashboard, select the Performance tab.

AI + machine learning

429 Throttling Detected on this resource

We observed that there have been 1,000 or more 429 throttling errors on this resource in a one day timeframe. Consider enabling autoscale to better handle higher call volumes and reduce the number of 429 errors.

Text Analytics Model Version Deprecation

Upgrade the model version to a newer model version or latest to utilize the latest and highest quality models.

Learn more about Cognitive Service - TAUpgradeToLatestModelVersion (Text Analytics Model Version Deprecation).

Text Analytics Model Version Deprecation

Upgrade the model version to a newer model version or latest to utilize the latest and highest quality models.

Learn more about Cognitive Service - TAUpgradeModelVersiontoLatest (Text Analytics Model Version Deprecation).

Upgrade to the latest Cognitive Service Text Analytics API version

Upgrade to the latest API version to get the best results in terms of model quality, performance and service availability. Also there are new features available as new endpoints starting from V3.0 such as personal data recognition, entity recognition and entity linking available as separate endpoints. In terms of changes in preview endpoints, we have Opinion Mining in SA endpoint, redacted text property in personal data endpoint

Learn more about Cognitive Service - UpgradeToLatestAPI (Upgrade to the latest Cognitive Service Text Analytics API version).

Upgrade to the latest API version of Azure Cognitive Service for Language

Upgrade to the latest API version to get the best results in terms of model quality, performance and service availability.

Learn more about Cognitive Service - UpgradeToLatestAPILanguage (Upgrade to the latest API version of Azure Cognitive Service for Language).

Upgrade to the latest Cognitive Service Text Analytics SDK version

Upgrade to the latest SDK version to get the best results in terms of model quality, performance and service availability. Also there are new features available as new endpoints starting from V3.0 such as personal data recognition, entity recognition and entity linking available as separate endpoints. In terms of changes in preview endpoints, we have Opinion Mining in SA endpoint, redacted text property in personal data endpoint

Learn more about Cognitive Service - UpgradeToLatestSDK (Upgrade to the latest Cognitive Service Text Analytics SDK version).

Upgrade to the latest Cognitive Service Language SDK version

Upgrade to the latest SDK version to get the best results in terms of model quality, performance and service availability.

Learn more about Cognitive Service - UpgradeToLatestSDKLanguage (Upgrade to the latest Cognitive Service Language SDK version).

Upgrade to the latest Azure AI Language SDK version

Learn more about Azure AI Language.

Analytics

Right-size Data Explorer resources for optimal performance.

This recommendation surfaces all Data Explorer resources that exceed the recommended data capacity (80%). The recommended action to improve the performance is to scale to the recommended configuration shown.

Learn more about Data explorer resource - Right-size ADX resource (Right-size Data Explorer resources for optimal performance.).

Review table cache policies for Data Explorer tables

This recommendation surfaces Data Explorer tables with a high number of queries that look back beyond the configured cache period (policy) - you see the top 10 tables by query percentage that access out-of-cache data. The recommended action to improve the performance: Limit queries on this table to the minimal necessary time range (within the defined policy). Alternatively, if data from the entire time range is required, increase the cache period to the recommended value.

Learn more about Data explorer resource - UpdateCachePoliciesForAdxTables (Review table cache policies for Data Explorer tables).

Reduce Data Explorer table cache policy for better performance

Reducing the table cache policy frees up unused data from the resource's cache and improves performance.

Learn more about Data explorer resource - ReduceCacheForAzureDataExplorerTablesToImprovePerformance (Reduce Data Explorer table cache policy for better performance).

Increase the cache in the cache policy

Based on your actual usage during the last month, update the cache policy to increase the hot cache for the table. The retention period must always be larger than the cache period. If you increase the cache and the retention period is lower than the cache period, update the retention policy. The analysis is based only on user queries that scanned data.

Learn more about Data explorer resource - IncreaseCacheForAzureDataExplorerTablesToImprovePerformance (Increase the cache in the cache policy).

Enable Optimized Autoscale for Data Explorer resources

Looks like your resource could have automatically scaled to improve performance (based on your actual usage during the last week, cache utilization, ingestion utilization, CPU, and streaming ingests utilization). To optimize costs and performance, we recommend enabling Optimized Autoscale.

Learn more about Data explorer resource - PerformanceEnableOptimizedAutoscaleAzureDataExplorer (Enable Optimized Autoscale for Data Explorer resources).

Reads happen on most recent data

More than 75% of your read requests are landing on the memstore, indicating that the reads are primarily on recent data. Recent data reads suggest that even if a flush happens on the memstore, the recent file needs to be accessed and put in the cache.

Learn more about HDInsight cluster - HBaseMemstoreReadPercentage (Reads happen on most recent data).

Consider using Accelerated Writes feature in your HBase cluster to improve cluster performance.

You're seeing this advisor recommendation because HDInsight team's system log shows that in the past seven days, your cluster has encountered the following scenarios:

High WAL sync time latency
High write request count (at least 3 one hour windows of over 1000 avg_write_requests/second/node)

These conditions are indicators that your cluster is suffering from high write latencies, which can be due to heavy workload on your cluster.

To improve the performance of your cluster, consider utilizing the Accelerated Writes feature provided by Azure HDInsight HBase. The Accelerated Writes feature for HDInsight Apache HBase clusters attaches premium SSD-managed disks to every RegionServer (worker node) instead of using cloud storage. As a result, it provides low write-latency and better resiliency for your applications.

To read more on this feature, visit link:

Learn more about HDInsight cluster - AccWriteCandidate (Consider using Accelerated Writes feature in your HBase cluster to improve cluster performance.).

More than 75% of your queries are full scan queries

More than 75% of the scan queries on your cluster are doing a full region/table scan. Modify your scan queries to avoid full region or table scans.

Learn more about HDInsight cluster - ScanQueryTuningcandidate (More than 75% of your queries are full scan queries.).

Check your region counts as you have blocking updates

Region counts needs to be adjusted to avoid updates getting blocked. It might require a scale up of the cluster by adding new nodes.

Learn more about HDInsight cluster - RegionCountCandidate (Check your region counts as you have blocking updates.).

Consider increasing the flusher threads

The flush queue size in your region servers is more than 100 or there are updates getting blocked frequently. Tuning of the flush handler is recommended.

Learn more about HDInsight cluster - FlushQueueCandidate (Consider increasing the flusher threads).

Consider increasing your compaction threads for compactions to complete faster

The compaction queue in your region servers is more than 2000 suggesting that more data requires compaction. Slower compactions can affect read performance as the number of files to read are more. More files without compaction can also affect the heap usage related to how files interact with Azure file system.

Learn more about HDInsight cluster - CompactionQueueCandidate (Consider increasing your compaction threads for compactions to complete faster).

Tables with Clustered Columnstore Indexes (CCI) with less than 60 million rows

Clustered columnstore tables are organized in data into segments. Having high segment quality is critical to achieving optimal query performance on a columnstore table. You can measure segment quality by the number of rows in a compressed row group.

Learn more about Synapse workspace - SynapseCCIGuidance (Tables with Clustered Columnstore Indexes (CCI) with less than 60 million rows).

Update SynapseManagementClient SDK Version

New SynapseManagementClient is using .NET SDK 4.0 or above.

Learn more about Synapse workspace - UpgradeSynapseManagementClientSDK (Update SynapseManagementClient SDK Version).

Compute

Use Managed disks to prevent disk I/O throttling

Your virtual machine disks belong to a storage account that has reached its scalability target, and is susceptible to I/O throttling. To protect your virtual machine from performance degradation and to simplify storage management, use Managed Disks.

Learn more about Virtual machine - ManagedDisksStorageAccount (Use Managed disks to prevent disk I/O throttling).

Convert Managed Disks from Standard HDD to Premium SSD for performance

We have noticed your Standard HDD disk is approaching performance targets. Azure premium SSDs deliver high-performance and low-latency disk support for virtual machines with IO-intensive workloads. Give your disk performance a boost by upgrading your Standard HDD disk to Premium SSD disk. Upgrading requires a VM reboot, which takes three to five minutes.

Learn more about Disk - MDHDDtoPremiumForPerformance (Convert Managed Disks from Standard HDD to Premium SSD for performance).

Enable Accelerated Networking to improve network performance and latency

We have detected that Accelerated Networking isn't enabled on VM resources in your existing deployment that might be capable of supporting this feature. If your VM OS image supports Accelerated Networking as detailed in the documentation, make sure to enable this free feature on these VMs to maximize the performance and latency of your networking workloads in cloud

Learn more about Virtual machine - AccelNetConfiguration (Enable Accelerated Networking to improve network performance and latency).

Use SSD Disks for your production workloads

We noticed that you're using SSD disks while also using Standard HDD disks on the same VM. Standard HDD managed disks are recommended for dev-test and backup; we recommend you use Premium SSDs or Standard SSDs for production. Premium SSDs deliver high-performance and low-latency disk support for virtual machines with IO-intensive workloads. Standard SSDs provide consistent and lower latency. Upgrade your disk configuration today for improved latency, reliability, and availability. Upgrading requires a VM reboot, which takes three to five minutes.

Learn more about Virtual machine - MixedDiskTypeToSSDPublic (Use SSD Disks for your production workloads).

Match production Virtual Machines with Production Disk for consistent performance and better latency

Production virtual machines need production disks if you want to get the best performance. We see that you're running a production level virtual machine, however, you're using a low performing disk with standard HDD. Upgrading disks that are attached to your production disks, either Standard SSD or Premium SSD, benefits you with a more consistent experience and improvements in latency.

Learn more about Virtual machine - MatchProdVMProdDisks (Match production Virtual Machines with Production Disk for consistent performance and better latency).

Accelerated Networking might require stopping and starting the VM

We have detected that Accelerated Networking isn't engaged on VM resources in your existing deployment even though the feature has been requested. In rare cases like this, it might be necessary to stop and start your VM, at your convenience, to re-engage AccelNet.

Learn more about Virtual machine - AccelNetDisengaged (Accelerated Networking might require stopping and starting the VM).

Take advantage of Ultra Disk low latency for your log disks and improve your database workload performance

Ultra disk is available in the same region as your database workload. Ultra disk offers high throughput, high IOPS, and consistent low latency disk storage for your database workloads: For Oracle DBs, you can now use either 4k or 512E sector sizes with Ultra disk depending on your Oracle DB version. For SQL server, using Ultra disk for your log disk might offer more performance for your database. See instructions here for migrating your log disk to Ultra disk.

Learn more about Virtual machine - AzureStorageVmUltraDisk (Take advantage of Ultra Disk low latency for your log disks and improve your database workload performance.).

Upgrade the size of your most active virtual machines to prevent resource exhaustion and improve performance

We analyzed data for the past seven days and identified virtual machines (VMs) with high utilization across different metrics (that is, CPU, Memory, and VM IO). Those VMs might experience performance issues since they're nearing or at their SKU's limits. Consider upgrading their SKU to improve performance.

Learn more about Virtual machine - UpgradeSizeHighVMUtilV0 (Upgrade the size of your most active virtual machines to prevent resource exhaustion and improve performance).

Containers

Unsupported Kubernetes version is detected

Unsupported Kubernetes version is detected. Ensure Kubernetes cluster runs with a supported version.

Learn more about Kubernetes service - UnsupportedKubernetesVersionIsDetected (Unsupported Kubernetes version is detected).

Unsupported Kubernetes version is detected

Unsupported Kubernetes version is detected. Ensure Kubernetes cluster runs with a supported version.

Learn more about HDInsight Cluster Pool - UnsupportedHiloAKSVersionIsDetected (Unsupported Kubernetes version is detected).

Clusters with a single node pool

We recommended that you add one or more node pools instead of using a single node pool. Multiple pools help to isolate critical system pods from your application to prevent misconfigured or rogue application pods from accidentally killing system pods.

Learn more about Kubernetes service - ClustersWithASingleNodePool (Clusters with a Single Node Pool).

Update Fleet API to the latest version

We have identified SDK calls from outdated Fleet API for resources under your subscription. We recommend switching to the latest SDK version, which ensures you receive the latest features and performance improvements.

Learn more about Kubernetes fleet manager | PREVIEW - UpdateToLatestFleetApi (Update Fleet API to the latest Version).

Databases

Configure your Azure Cosmos DB query page size (MaxItemCount) to -1

You're using the query page size of 100 for queries for your Azure Cosmos DB container. We recommend using a page size of -1 for faster scans.

Learn more about Azure Cosmos DB account - CosmosDBQueryPageSize (Configure your Azure Cosmos DB query page size (MaxItemCount) to -1).

Add composite indexes to your Azure Cosmos DB container

Your Azure Cosmos DB containers are running ORDER BY queries incurring high Request Unit (RU) charges. It's recommended to add composite indexes to your containers' indexing policy to improve the RU consumption and decrease the latency of these queries.

Learn more about Azure Cosmos DB account - CosmosDBOrderByHighRUCharge (Add composite indexes to your Azure Cosmos DB container).

Optimize your Azure Cosmos DB indexing policy to only index what's needed

Your Azure Cosmos DB containers are using the default indexing policy, which indexes every property in your documents. Because you're storing large documents, a high number of properties get indexed, resulting in high Request Unit consumption and poor write latency. To optimize write performance, we recommend overriding the default indexing policy to only index the properties used in your queries.

Learn more about Azure Cosmos DB account - CosmosDBDefaultIndexingWithManyPaths (Optimize your Azure Cosmos DB indexing policy to only index what's needed).

Use hierarchical partition keys for optimal data distribution

Your account has a custom setting that allows the logical partition size in a container to exceed the limit of 20 GB. The Azure Cosmos DB team applied this setting as a temporary measure to give you time to rearchitect your application with a different partition key. It isn't recommended as a long-term solution, as SLA guarantees aren't honored when the limit is increased. You can now use hierarchical partition keys (preview) to rearchitect your application. The feature allows you to exceed the 20-GB limit by setting up to three partition keys, ideal for multitenant scenarios or workloads that use synthetic keys.

Learn more about Azure Cosmos DB account - CosmosDBHierarchicalPartitionKey (Use hierarchical partition keys for optimal data distribution).

Enhance Performance by Scaling Up for Optimal Resource Utilization

Maximizing the efficiency of your system's resources is crucial for maintaining top-notch performance. Our system closely monitors CPU usage, and when it crosses the 90% threshold over a 12-hour period, a proactive alert is triggered. This alert not only informs Azure Cosmos DB for MongoDB vCore users of the elevated CPU consumption but also provides valuable guidance on scaling up to a higher tier. By upgrading to a more robust tier, you can unlock improved performance and ensure your system operates at its peak potential.

Learn more about Scaling and configuring Your Azure Cosmos DB for MongoDB vCore cluster.

PerformanceBoostervCore

When CPU usage surpasses 90% within a 12-hour timeframe, users are notified about the high usage. Additionally it advises them to scale up to a higher tier to get a better performance.

Learn more about Cosmos DB account - ScaleUpvCoreRecommendation (PerformanceBoostervCore).

Scale the storage limit for MariaDB server

Our system shows that the server might be constrained because it's approaching limits for the currently provisioned storage values. Approaching the storage limits might result in degraded performance or the server moved to read-only mode. To ensure continued performance, we recommend increasing the provisioned storage amount or turning ON the "Auto-Growth" feature for automatic storage increases

Learn more about MariaDB server - OrcasMariaDbStorageLimit (Scale the storage limit for MariaDB server).

Increase the MariaDB server vCores

Our system shows that the CPU has been running under high utilization for an extended time period over the last seven days. High CPU utilization might lead to slow query performance. To improve performance, we recommend moving to a larger compute size.

Learn more about MariaDB server - OrcasMariaDbCpuOverload (Increase the MariaDB server vCores).

Scale the MariaDB server to higher SKU

Our system shows that the server might be unable to support the connection requests because of the maximum supported connections for the given SKU, which might result in a large number of failed connections requests which adversely affect the performance. To improve performance, we recommend moving to higher memory SKU by increasing vCore or switching to Memory-Optimized SKUs.

Learn more about MariaDB server - OrcasMariaDbConcurrentConnection (Scale the MariaDB server to higher SKU).

Move your MariaDB server to Memory Optimized SKU

Our system shows that there is high churn in the buffer pool for this server which can result in slower query performance and increased IOPS. To improve performance, review your workload queries to identify opportunities to minimize memory consumed. If no such opportunity found, then we recommend moving to higher SKU with more memory or increase storage size to get more IOPS.

Learn more about MariaDB server - OrcasMariaDbMemoryCache (Move your MariaDB server to Memory Optimized SKU).

Increase the reliability of audit logs

Our system shows that the server's audit logs might have been lost over the past day. Lost audit logs can occur when your server is experiencing a CPU-heavy workload, or a server generates a large number of audit logs over a short time period. We recommend only logging the necessary events required for your audit purposes using the following server parameters: audit_log_events, audit_log_exclude_users, audit_log_include_users. If the CPU usage on your server is high due to your workload, we recommend increasing the server's vCores to improve performance.

Learn more about MariaDB server - OrcasMariaDBAuditLog (Increase the reliability of audit logs).

Scale the storage limit for MySQL server

Our system shows that the server might be constrained because it is approaching limits for the currently provisioned storage values. Approaching the storage limits might result in degraded performance or in the server being moved to read-only mode. To ensure continued performance, we recommend increasing the provisioned storage amount or turning ON the "Auto-Growth" feature for automatic storage increases

Learn more about MySQL server - OrcasMySQLStorageLimit (Scale the storage limit for MySQL server).

Scale the MySQL server to higher SKU

Our system shows that the server might be unable to support the connection requests because of the maximum supported connections for the given SKU, which might result in a large number of failed connections requests that adversely affect the performance. To improve performance, we recommend moving to a higher memory SKU by increasing vCore or switching to Memory-Optimized SKUs.

Learn more about MySQL server - OrcasMySQLConcurrentConnection (Scale the MySQL server to higher SKU).

Increase the MySQL server vCores

Learn more about MySQL server - OrcasMySQLCpuOverload (Increase the MySQL server vCores).

Move your MySQL server to Memory Optimized SKU

Learn more about MySQL server - OrcasMySQLMemoryCache (Move your MySQL server to Memory Optimized SKU).

Add a MySQL Read Replica server

Our system shows that you might have a read intensive workload running, which results in resource contention for this server. Resource contention might lead to slow query performance for the server. To improve performance, we recommend you add a read replica, and offload some of your read workloads to the replica.

Learn more about MySQL server - OrcasMySQLReadReplica (Add a MySQL Read Replica server).

Improve MySQL connection management

Our system shows that your application connecting to MySQL server might be managing connections poorly, which might result in unnecessary resource consumption and overall higher application latency. To improve connection management, we recommend that you reduce the number of short-lived connections and eliminate unnecessary idle connections. You can do this by configuring a server side connection-pooler, such as ProxySQL.

Learn more about MySQL server - OrcasMySQLConnectionPooling (Improve MySQL connection management).

Increase the reliability of audit logs

Our system shows that the server's audit logs might have been lost over the past day. This can occur when your server is experiencing a CPU heavy workload or a server generates a large number of audit logs over a short time period. We recommend only logging the necessary events required for your audit purposes using the following server parameters: audit_log_events, audit_log_exclude_users, audit_log_include_users. If the CPU usage on your server is high due to your workload, we recommend increasing the server's vCores to improve performance.

Learn more about MySQL server - OrcasMySQLAuditLog (Increase the reliability of audit logs).

Improve performance by optimizing MySQL temporary-table sizing

Our system shows that your MySQL server might be incurring unnecessary I/O overhead due to low temporary-table parameter settings. This might result in unnecessary disk-based transactions and reduced performance. We recommend that you increase the 'tmp_table_size' and 'max_heap_table_size' parameter values to reduce the number of disk-based transactions.

Learn more about MySQL server - OrcasMySqlTmpTables (Improve performance by optimizing MySQL temporary-table sizing).

Improve MySQL connection latency

Our system shows that your application connecting to MySQL server might be managing connections poorly. This might result in higher application latency. To improve connection latency, we recommend that you enable connection redirection. This can be done by enabling the connection redirection feature of the PHP driver.

Learn more about MySQL server - OrcasMySQLConnectionRedirection (Improve MySQL connection latency).

Increase the storage limit for MySQL Flexible Server

Learn more about Azure Database for MySQL flexible server - OrcasMeruMySqlStorageUpsell (Increase the storage limit for MySQL Flexible Server).

Scale the MySQL Flexible Server to a higher SKU

Our system shows that your Flexible Server is exceeding the connection limits associated with your current SKU. A large number of failed connection requests might adversely affect server performance. To improve performance, we recommend increasing the number of vCores or switching to a higher SKU.

Learn more about Azure Database for MySQL flexible server - OrcasMeruMysqlConnectionUpsell (Scale the MySQL Flexible Server to a higher SKU).

Increase the MySQL Flexible Server vCores.

Learn more about Azure Database for MySQL flexible server - OrcasMeruMysqlCpuUpcell (Increase the MySQL Flexible Server vCores.).

Improve performance by optimizing MySQL temporary-table sizing.

Our system shows that your MySQL server might be incurring unnecessary I/O overhead due to low temporary-table parameter settings. Unnecessary I/O overhead might result in unnecessary disk-based transactions and reduced performance. We recommend that you increase the 'tmp_table_size' and 'max_heap_table_size' parameter values to reduce the number of disk-based transactions.

Learn more about Azure Database for MySQL flexible server - OrcasMeruMysqlTmpTable (Improve performance by optimizing MySQL temporary-table sizing.).

Move your MySQL server to Memory Optimized SKU

Our system shows that there is high memory usage for this server that can result in slower query performance and increased IOPS. To improve performance, review your workload queries to identify opportunities to minimize memory consumed. If no such opportunity found, then we recommend moving to higher SKU with more memory or increase storage size to get more IOPS.

Learn more about Azure Database for MySQL flexible server - OrcasMeruMysqlMemoryUpsell (Move your MySQL server to Memory Optimized SKU).

Add a MySQL Read Replica server

Our system shows that you might have a read intensive workload running, which results in resource contention for this server. This might lead to slow query performance for the server. To improve performance, we recommend you add a read replica, and offload some of your read workloads to the replica.

Learn more about Azure Database for MySQL flexible server - OrcasMeruMysqlReadReplicaUpsell (Add a MySQL Read Replica server).

Increase the work_mem to avoid excessive disk spilling from sort and hash

Our system shows that the configuration work_mem is too small for your PostgreSQL server which is resulting in disk spilling and degraded query performance. To improve this, we recommend increasing the work_mem limit for the server, which helps to reduce the scenarios when the sort or hash happens on disk and improves the overall query performance.

Learn more about PostgreSQL server - OrcasPostgreSqlWorkMem (Increase the work_mem to avoid excessive disk spilling from sort and hash).

Boost your workload performance by 30% with the new Ev5 compute hardware

With the new Ev5 compute hardware, you can boost workload performance by 30% with higher concurrency and better throughput. Navigate to the Compute+Storage option on the Azure portal and switch to Ev5 compute at no extra cost. Ev5 compute provides best performance among other VM series in terms of QPS and latency.

Learn more about Azure Database for MySQL flexible server - OrcasMeruMySqlComputeSeriesUpgradeEv5 (Boost your workload performance by 30% with the new Ev5 compute hardware).

Scale the storage limit for PostgreSQL server

Learn more about PostgreSQL server - OrcasPostgreSqlStorageLimit (Scale the storage limit for PostgreSQL server).

Scale the PostgreSQL server to higher SKU

Our system shows that the server might be unable to support the connection requests because of the maximum supported connections for the given SKU, which might result in a large number of failed connections requests adversely affecting performance. To improve performance, we recommend moving to higher memory SKU by increasing vCore or switching to Memory-Optimized SKUs.

Learn more about PostgreSQL server - OrcasPostgreSqlConcurrentConnection (Scale the PostgreSQL server to higher SKU).

Move your PostgreSQL server to Memory Optimized SKU

Learn more about PostgreSQL server - OrcasPostgreSqlMemoryCache (Move your PostgreSQL server to Memory Optimized SKU).

Add a PostgreSQL Read Replica server

Our system shows that you might have a read intensive workload running, which results in resource contention for this server. Resource contention can lead to slow query performance for the server. To improve performance, we recommend you add a read replica, and offload some of your read workloads to the replica.

Learn more about PostgreSQL server - OrcasPostgreSqlReadReplica (Add a PostgreSQL Read Replica server).

Increase the PostgreSQL server vCores

Learn more about PostgreSQL server - OrcasPostgreSqlCpuOverload (Increase the PostgreSQL server vCores).

Improve PostgreSQL connection management

Our system shows that your PostgreSQL server might not be managing connections efficiently, which can result in unnecessary resource consumption and overall higher application latency. To improve connection management, we recommend that you reduce the number of short-lived connections and eliminate unnecessary idle connections by configuring a server side connection-pooler, such as PgBouncer.

Learn more about PostgreSQL server - OrcasPostgreSqlConnectionPooling (Improve PostgreSQL connection management).

Improve PostgreSQL log performance

Our system shows that your PostgreSQL server has been configured to output VERBOSE error logs. This setting can be useful for troubleshooting your database, but it can also result in reduced database performance. To improve performance, we recommend that you change the log_error_verbosity parameter to the DEFAULT setting.

Learn more about PostgreSQL server - OrcasPostgreSqlLogErrorVerbosity (Improve PostgreSQL log performance).

Optimize query statistics collection on an Azure Database for PostgreSQL

Our system shows that your PostgreSQL server has been configured to track query statistics using the pg_stat_statements module. While useful for troubleshooting, it can also result in reduced server performance. To improve performance, we recommend that you change the pg_stat_statements.track parameter to NONE.

Learn more about PostgreSQL server - OrcasPostgreSqlStatStatementsTrack (Optimize query statistics collection on an Azure Database for PostgreSQL).

Optimize query store on an Azure Database for PostgreSQL when not troubleshooting

Our system shows that your PostgreSQL database has been configured to track query performance using the pg_qs.query_capture_mode parameter. While troubleshooting, we suggest setting the pg_qs.query_capture_mode parameter to TOP or ALL. When not troubleshooting, we recommend that you set the pg_qs.query_capture_mode parameter to NONE.

Learn more about PostgreSQL server - OrcasPostgreSqlQueryCaptureMode (Optimize query store on an Azure Database for PostgreSQL when not troubleshooting).