Slow query logs in Azure Database for MariaDB

Important

Azure Database for MariaDB is on the retirement path. We strongly recommend that you migrate to Azure Database for MySQL. For more information about migrating to Azure Database for MySQL, see What's happening to Azure Database for MariaDB?.

In Azure Database for MariaDB, the slow query log is available to users. Access to the transaction log is not supported. The slow query log can be used to identify performance bottlenecks for troubleshooting.

For more information about the slow query log, see the MariaDB documentation for slow query log.

When Query Store is enabled on your server, you may see the queries like "CALL mysql.az_procedure_collect_wait_stats (900, 30);" logged in your slow query logs. This behavior is expected as the Query Store feature collects statistics about your queries.

Configure slow query logging

By default the slow query log is disabled. To enable it, set slow_query_log to ON. This can be enabled using the Azure portal or Azure CLI.

Other parameters you can adjust include:

  • long_query_time: if a query takes longer than long_query_time (in seconds) that query is logged. The default is 10 seconds.
  • log_slow_admin_statements: if ON includes administrative statements like ALTER_TABLE and ANALYZE_TABLE in the statements written to the slow_query_log.
  • log_queries_not_using_indexes: determines whether queries that do not use indexes are logged to the slow_query_log
  • log_throttle_queries_not_using_indexes: This parameter limits the number of non-index queries that can be written to the slow query log. This parameter takes effect when log_queries_not_using_indexes is set to ON.
  • log_output: if "File", allows the slow query log to be written to both the local server storage and to Azure Monitor Diagnostic Logs. If "None", the slow query log will only be written to Azure Monitor Diagnostics Logs.

Important

If your tables are not indexed, setting the log_queries_not_using_indexes and log_throttle_queries_not_using_indexes parameters to ON may affect MariaDB performance since all queries running against these non-indexed tables will be written to the slow query log.

If you plan on logging slow queries for an extended period of time, it is recommended to set log_output to "None". If set to "File", these logs are written to the local server storage and can affect MariaDB performance.

See the MariaDB slow query log documentation for full descriptions of the slow query log parameters.

Access slow query logs

There are two options for accessing slow query logs in Azure Database for MariaDB: local server storage or Azure Monitor Diagnostic Logs. This is set using the log_output parameter.

For local server storage, you can list and download slow query logs using the Azure portal or the Azure CLI. In the Azure portal, navigate to your server in the Azure portal. Under the Monitoring heading, select the Server Logs page. For more information on Azure CLI, see Configure and access server logs using Azure CLI.

Azure Monitor Diagnostic Logs allows you to pipe slow query logs to Azure Monitor Logs (Log Analytics), Azure Storage, or Event Hubs. See below for more information.

Local server storage log retention

When logging to the server's local storage, logs are available for up to seven days from their creation. If the total size of the available logs exceeds 7 GB, then the oldest files are deleted until space is available. The 7 GB storage limit for the server logs is available free of cost and cannot be extended.

Logs are rotated every 24 hours or 7 GB, whichever comes first.

Note

The above log retention does not apply to logs that are piped using Azure Monitor Diagnostic Logs. You can change the retention period for the data sinks being emitted to (ex. Azure Storage).

Diagnostic logs

Azure Database for MariaDB is integrated with Azure Monitor Diagnostic Logs. Once you have enabled slow query logs on your MariaDB server, you can choose to have them emitted to Azure Monitor logs, Event Hubs, or Azure Storage. To learn more about how to enable diagnostic logs, see the how to section of the diagnostic logs documentation.

The following table describes what's in each log. Depending on the output method, the fields included and the order in which they appear may vary.

Property Description
TenantId Your tenant ID
SourceSystem Azure
TimeGenerated [UTC] Time stamp when the log was recorded in UTC
Type Type of the log. Always AzureDiagnostics
SubscriptionId GUID for the subscription that the server belongs to
ResourceGroup Name of the resource group the server belongs to
ResourceProvider Name of the resource provider. Always MICROSOFT.DBFORMARIADB
ResourceType Servers
ResourceId Resource URI
Resource Name of the server
Category MySqlSlowLogs
OperationName LogEvent
Logical_server_name_s Name of the server
start_time_t [UTC] Time the query began
query_time_s Total time the query took to execute
lock_time_s Total time the query was locked
user_host_s Username
rows_sent_s Number of rows sent
rows_examined_s Number of rows examined
last_insert_id_s last_insert_id
insert_id_s Insert ID
sql_text_s Full query
server_id_s Server ID
thread_id_s Thread ID
\_ResourceId Resource URI

Note

For sql_text, log will be truncated if it exceeds 2048 characters.

Analyze logs in Azure Monitor Logs

Once your slow query logs are piped to Azure Monitor Logs through Diagnostic Logs, you can perform further analysis of your slow queries. Below are some sample queries to help you get started. Make sure to update the below with your server name.

  • Queries longer than 10 seconds on a particular server

    AzureDiagnostics
    | where LogicalServerName_s == '<your server name>'
    | where Category == 'MySqlSlowLogs'
    | project TimeGenerated, LogicalServerName_s, event_class_s, start_time_t , query_time_d, sql_text_s 
    | where query_time_d > 10
    
  • List top 5 longest queries on a particular server

    AzureDiagnostics
    | where LogicalServerName_s == '<your server name>'
    | where Category == 'MySqlSlowLogs'
    | project TimeGenerated, LogicalServerName_s, event_class_s, start_time_t , query_time_d, sql_text_s 
    | order by query_time_d desc
    | take 5
    
  • Summarize slow queries by minimum, maximum, average, and standard deviation query time on a particular server

    AzureDiagnostics
    | where LogicalServerName_s == '<your server name>'
    | where Category == 'MySqlSlowLogs'
    | project TimeGenerated, LogicalServerName_s, event_class_s, start_time_t , query_time_d, sql_text_s 
    | summarize count(), min(query_time_d), max(query_time_d), avg(query_time_d), stdev(query_time_d), percentile(query_time_d, 95) by LogicalServerName_s
    
  • Graph the slow query distribution on a particular server

    AzureDiagnostics
    | where LogicalServerName_s == '<your server name>'
    | where Category == 'MySqlSlowLogs'
    | project TimeGenerated, LogicalServerName_s, event_class_s, start_time_t , query_time_d, sql_text_s 
    | summarize count() by LogicalServerName_s, bin(TimeGenerated, 5m)
    | render timechart
    
  • Display queries longer than 10 seconds across all MariaDB servers with Diagnostic Logs enabled

    AzureDiagnostics
    | where Category == 'MySqlSlowLogs'
    | project TimeGenerated, LogicalServerName_s, event_class_s, start_time_t , query_time_d, sql_text_s 
    | where query_time_d > 10
    

Next Steps