Troubleshoot issues with advanced diagnostics queries with Azure Cosmos DB for Apache Gremlin

Article
12/09/2022

APPLIES TO: NoSQL MongoDB Cassandra Gremlin

In this article, we'll cover how to write more advanced queries to help troubleshoot issues with your Azure Cosmos DB account by using diagnostics logs sent to Azure Diagnostics (legacy) and resource-specific (preview) tables.

For Azure Diagnostics tables, all data is written into one single table. Users specify which category they want to query. If you want to view the full-text query of your request, see Monitor Azure Cosmos DB data by using diagnostic settings in Azure to learn how to enable this feature.

For resource-specific tables, data is written into individual tables for each category of the resource. We recommend this mode because it:

Makes it much easier to work with the data.
Provides better discoverability of the schemas.
Improves performance across both ingestion latency and query times.

Common queries

Common queries are shown in the resource-specific and Azure Diagnostics tables.

Top N(10) Request Unit (RU) consuming requests or queries in a specific time frame

Resource-specific
Azure Diagnostics

let topRequestsByRUcharge = CDBDataPlaneRequests 
| where TimeGenerated > ago(24h)
| project  RequestCharge , TimeGenerated, ActivityId;
CDBGremlinRequests
| project PIICommandText, ActivityId, DatabaseName , CollectionName
| join kind=inner topRequestsByRUcharge on ActivityId
| project DatabaseName , CollectionName , PIICommandText , RequestCharge, TimeGenerated
| order by RequestCharge desc
| take 10

let topRequestsByRUcharge = AzureDiagnostics
| where Category == "DataPlaneRequests" and TimeGenerated > ago(1h)
| project  requestCharge_s , TimeGenerated, activityId_g;
AzureDiagnostics
| where Category == "GremlinRequests"
| project piiCommandText_s, activityId_g, databasename_s , collectionname_s
| join kind=inner topRequestsByRUcharge on activityId_g
| project databasename_s , collectionname_s , piiCommandText_s , requestCharge_s, TimeGenerated
| order by requestCharge_s desc
| take 10

Requests throttled (statusCode = 429) in a specific time window

Resource-specific
Azure Diagnostics

let throttledRequests = CDBDataPlaneRequests
| where StatusCode == "429"
| project  OperationName , TimeGenerated, ActivityId;
CDBGremlinRequests
| project PIICommandText, ActivityId, DatabaseName , CollectionName
| join kind=inner throttledRequests on ActivityId
| project DatabaseName , CollectionName , PIICommandText , OperationName, TimeGenerated

let throttledRequests = AzureDiagnostics
| where Category == "DataPlaneRequests"
| where statusCode_s == "429"
| project  OperationName , TimeGenerated, activityId_g;
AzureDiagnostics
| where Category == "GremlinRequests"
| project piiCommandText_s, activityId_g, databasename_s , collectionname_s
| join kind=inner throttledRequests on activityId_g
| project databasename_s , collectionname_s , piiCommandText_s , OperationName, TimeGenerated

Queries with large response lengths (payload size of the server response)

Resource-specific
Azure Diagnostics

let operationsbyUserAgent = CDBDataPlaneRequests
| project OperationName, DurationMs, RequestCharge, ResponseLength, ActivityId;
CDBGremlinRequests
//specify collection and database
 //| where DatabaseName == "DB NAME" and CollectionName == "COLLECTIONNAME"
| join kind=inner operationsbyUserAgent on ActivityId
| summarize max(ResponseLength) by PIICommandText
| order by max_ResponseLength desc

let operationsbyUserAgent = AzureDiagnostics
 | where Category=="DataPlaneRequests"
| project OperationName, duration_s, requestCharge_s, responseLength_s, activityId_g;
AzureDiagnostics
| where Category == "GremlinRequests"
//| where databasename_s == "DB NAME" and collectioname_s == "COLLECTIONNAME"
| join kind=inner operationsbyUserAgent on activityId_g
| summarize max(responseLength_s1) by piiCommandText_s
| order by max_responseLength_s1 desc

RU consumption by physical partition (across all replicas in the replica set)

Resource-specific
Azure Diagnostics

CDBPartitionKeyRUConsumption
| where TimeGenerated >= now(-1d)
//specify collection and database
//| where DatabaseName == "DB NAME" and CollectionName == "COLLECTIONNAME"
// filter by operation type
//| where operationType_s == 'Create'
| summarize sum(todouble(RequestCharge)) by toint(PartitionKeyRangeId)
| render columnchart

AzureDiagnostics
| where TimeGenerated >= now(-1d)
| where Category == 'PartitionKeyRUConsumption'
//specify collection and database
//| where databasename_s == "DB NAME" and collectioname_s == "COLLECTIONNAME"
// filter by operation type
//| where operationType_s == 'Create'
| summarize sum(todouble(requestCharge_s)) by toint(partitionKeyRangeId_s)
| render columnchart

RU consumption by logical partition (across all replicas in the replica set)

Resource-specific
Azure Diagnostics

CDBPartitionKeyRUConsumption
| where TimeGenerated >= now(-1d)
//specify collection and database
//| where DatabaseName == "DB NAME" and CollectionName == "COLLECTIONNAME"
// filter by operation type
//| where operationType_s == 'Create'
| summarize sum(todouble(RequestCharge)) by PartitionKey, PartitionKeyRangeId
| render columnchart

AzureDiagnostics
| where TimeGenerated >= now(-1d)
| where Category == 'PartitionKeyRUConsumption'
//specify collection and database
//| where databasename_s == "DB NAME" and collectioname_s == "COLLECTIONNAME"
// filter by operation type
//| where operationType_s == 'Create'
| summarize sum(todouble(requestCharge_s)) by partitionKey_s, partitionKeyRangeId_s
| render columnchart

Next steps

For more information on how to create diagnostic settings for Azure Cosmos DB, see Create diagnostic settings.
For detailed information about how to create a diagnostic setting by using the Azure portal, the Azure CLI, or PowerShell, see Create diagnostic settings to collect platform logs and metrics in Azure.