Monitoring OpenSearch logs with Amazon CloudWatch Logs
Amazon OpenSearch Service exposes the following OpenSearch logs through Amazon CloudWatch Logs:
-
Error logs
Search shard slow logs, indexing shard slow logs, and error logs are useful for troubleshooting
performance and stability issues. Audit logs track user activity for compliance purposes. All
the logs are disabled by default. If enabled, standard CloudWatch pricing
Note
Error logs are available only for OpenSearch and Elasticsearch versions 5.1 and later. Slow logs are available for all OpenSearch and Elasticsearch versions.
For its logs, OpenSearch uses Apache
Log4j 2TRACE
, DEBUG
, INFO
, WARN
,
ERROR
, and FATAL
.
If you enable error logs, OpenSearch Service publishes log lines of WARN
,
ERROR
, and FATAL
to CloudWatch. OpenSearch Service also publishes several
exceptions from the DEBUG
level, including the following:
-
org.opensearch.index.mapper.MapperParsingException
-
org.opensearch.index.query.QueryShardException
-
org.opensearch.action.search.SearchPhaseExecutionException
-
org.opensearch.common.util.concurrent.OpenSearchRejectedExecutionException
-
java.lang.IllegalArgumentException
Error logs can help with troubleshooting in many situations, including the following:
-
Painless script compilation issues
-
Invalid queries
-
Indexing issues
-
Snapshot failures
-
Index State Management migration failures
Note
OpenSearch Service does not log all errors that occur.
Topics
Enabling log publishing (console)
The OpenSearch Service console is the simplest way to enable the publishing of logs to CloudWatch.
To enable log publishing to CloudWatch (console)
-
Go to https://aws.amazon.com
, and then choose Sign In to the Console. -
Under Analytics, choose Amazon OpenSearch Service.
-
Select the domain you want to update.
-
On the Logs tab, select a log type and choose Enable.
-
Create a new CloudWatch log group or choose an existing one.
Note
If you plan to enable multiple logs, we recommend publishing each to its own log group. This separation makes the logs easier to scan.
-
Choose an access policy that contains the appropriate permissions, or create a policy using the JSON that the console provides:
{ "Version": "2012-10-17", "Statement": [ { "Effect": "Allow", "Principal": { "Service": "es.amazonaws.com" }, "Action": [ "logs:PutLogEvents", "logs:CreateLogStream" ], "Resource": "
cw_log_group_arn
:*" } ] }We recommend that you add the
aws:SourceAccount
andaws:SourceArn
condition keys to the policy to protect yourself against the confused deputy problem. The source account is the owner of the domain and the source ARN is the ARN of the domain. Your domain must be on service software R20211203 or later in order to add these condition keys.For example, you could add the following condition block to the policy:
"Condition": { "StringEquals": { "aws:SourceAccount": "
account-id
" }, "ArnLike": { "aws:SourceArn": "arn:aws:es:region
:account-id
:domain/domain-name
" } }Important
CloudWatch Logs supports 10 resource policies per Region. If you plan to enable logs for several OpenSearch Service domains, you should create and reuse a broader policy that includes multiple log groups to avoid reaching this limit. For steps on updating your policy, see Enabling log publishing (AWS CLI).
-
Choose Enable.
The status of your domain changes from Active to Processing. The status must return to Active before log publishing is enabled. This change typically takes 30 minutes, but can take longer depending on your domain configuration.
If you enabled one of the shard slow logs, see Setting shard slow log thresholds. If you enabled audit logs, see Step 2: Turn on audit logs in OpenSearch Dashboards. If you enabled only error logs, you don't need to perform any additional configuration steps.
Enabling log publishing (AWS CLI)
Before you can enable log publishing, you need a CloudWatch log group. If you don't already have one, you can create one using the following command:
aws logs create-log-group --log-group-name
my-log-group
Enter the next command to find the log group's ARN, and then make a note of it:
aws logs describe-log-groups --log-group-name
my-log-group
Now you can give OpenSearch Service permissions to write to the log group. You must provide the log group's ARN near the end of the command:
aws logs put-resource-policy \ --policy-name my-policy \ --policy-document '{ "Version": "2012-10-17", "Statement": [{ "Sid": "", "Effect": "Allow", "Principal": { "Service": "es.amazonaws.com"}, "Action":[ "logs:PutLogEvents","logs:CreateLogStream"],"Resource": "
cw_log_group_arn
:*"}]}'
Important
CloudWatch Logs supports 10 resource policies per Region. If you plan to enable shard slow logs for several OpenSearch Service domains, you should create and reuse a broader policy that includes multiple log groups to avoid reaching this limit.
If you need to review this policy at a later time, use the aws logs
describe-resource-policies
command. To update the policy, issue the same aws
logs put-resource-policy
command with a new policy document.
Finally, you can use the --log-publishing-options
option to enable
publishing. The syntax for the option is the same for both the create-domain
and
update-domain-config
commands.
Parameter | Valid Values |
---|---|
--log-publishing-options |
SEARCH_SLOW_LOGS={CloudWatchLogsLogGroupArn= |
INDEX_SLOW_LOGS={CloudWatchLogsLogGroupArn= |
|
ES_APPLICATION_LOGS={CloudWatchLogsLogGroupArn= |
|
AUDIT_LOGS={CloudWatchLogsLogGroupArn= |
Note
If you plan to enable multiple logs, we recommend publishing each to its own log group. This separation makes the logs easier to scan.
Example
The following example enables the publishing of search and indexing shard slow logs for the specified domain:
aws opensearch update-domain-config \ --domain-name my-domain \ --log-publishing-options "SEARCH_SLOW_LOGS={CloudWatchLogsLogGroupArn=arn:aws:logs:us-east-1:123456789012:log-group:my-log-group,Enabled=true},INDEX_SLOW_LOGS={CloudWatchLogsLogGroupArn=arn:aws:logs:us-east-1:123456789012:log-group:my-other-log-group,Enabled=true}"
To disable publishing to CloudWatch, run the same command with
Enabled=false
.
If you enabled one of the shard slow logs, see Setting shard slow log thresholds. If you enabled audit logs, see Step 2: Turn on audit logs in OpenSearch Dashboards. If you enabled only error logs, you don't need to perform any additional configuration steps.
Enabling log publishing (AWS SDKs)
Before you can enable log publishing, you must first create a CloudWatch log group, get its ARN, and give OpenSearch Service permissions to write to it. The relevant operations are documented in the Amazon CloudWatch Logs API Reference:
-
CreateLogGroup
-
DescribeLogGroup
-
PutResourcePolicy
You can access these operations using the AWS SDKs
The AWS SDKs (except the Android and iOS SDKs) support all the operations that are
defined in the Amazon OpenSearch Service API Reference, including the
--log-publishing-options
option for CreateDomain
and
UpdateDomainConfig
.
If you enabled one of the shard slow logs, see Setting shard slow log thresholds. If you enabled only error logs, you don't need to perform any additional configuration steps.
Enabling log publishing (CloudFormation)
In this example, we use CloudFormation to create a log group called
opensearch-logs
, assign the appropriate permissions, and then create a domain
with log publishing enabled for application logs, search shard slow logs, and indexing slow
logs.
Before you can enable log publishing, you need to create a CloudWatch log group:
Resources: OpenSearchLogGroup: Type: AWS::Logs::LogGroup Properties: LogGroupName:
opensearch-logs
Outputs: Arn: Value: 'Fn::GetAtt': - OpenSearchLogGroup - Arn
The template outputs the ARN of the log group. In this case, the ARN is
arn:aws:logs:us-east-1:123456789012:log-group:opensearch-logs
.
Using the ARN, create a resource policy that gives OpenSearch Service permissions to write to the log group:
Resources: OpenSearchLogPolicy: Type: AWS::Logs::ResourcePolicy Properties: PolicyName:
my-policy
PolicyDocument: "{ \"Version\": \"2012-10-17\", \"Statement\": [{ \"Sid\": \"\", \"Effect\": \"Allow\", \"Principal\": { \"Service\": \"es.amazonaws.com\"}, \"Action\":[ \"logs:PutLogEvents\",\"logs:CreateLogStream\"],\"Resource\": \"arn:aws:logs:us-east-1:123456789012:log-group:opensearch-logs:*\"}]}"
Finally, create the following CloudFormation stack, which generates an OpenSearch Service domain with log publishing. The access policy permits the user for the AWS account to make all HTTP requests to the domain.
Resources: OpenSearchServiceDomain: Type: "AWS::OpenSearchService::Domain" Properties: DomainName:
my-domain
EngineVersion: "OpenSearch_1.0" ClusterConfig: InstanceCount: 2 InstanceType: "r6g.xlarge.search" DedicatedMasterEnabled: true DedicatedMasterCount: 3 DedicatedMasterType: "r6g.xlarge.search" EBSOptions: EBSEnabled: true VolumeSize: 10 VolumeType: "gp2" AccessPolicies: Version: "2012-10-17" Statement: Effect: "Allow" Principal: AWS: "arn:aws:iam::123456789012
:user/es-user" Action: "es:*" Resource: "arn:aws:es:us-east-1:123456789012
:domain/my-domain
/*" LogPublishingOptions: ES_APPLICATION_LOGS: CloudWatchLogsLogGroupArn: "arn:aws:logs:us-east-1:123456789012
:log-group:opensearch-logs" Enabled: true SEARCH_SLOW_LOGS: CloudWatchLogsLogGroupArn: "arn:aws:logs:us-east-1:123456789012
:log-group:opensearch-logs" Enabled: true INDEX_SLOW_LOGS: CloudWatchLogsLogGroupArn: "arn:aws:logs:us-east-1:123456789012
:log-group:opensearch-logs" Enabled: true
For detailed syntax information, see the log publishing options in the AWS CloudFormation User Guide.
Setting search request slow log thresholds
Search request slow logs
You can specify search request slow logs with cluster settings. This differs from shard slow logs, which you enable with index settings. For example, you can specify the following settings through the OpenSearch REST API:
PUT
domain-endpoint
/_cluster/settings { "transient": { "cluster.search.request.slowlog.threshold.warn": "5s", "cluster.search.request.slowlog.threshold.info": "2s" } }
Setting shard slow log thresholds
OpenSearch disables shard slow logs
For example, you can specify these settings through the OpenSearch REST API:
PUT
domain-endpoint
/index
/_settings { "index.search.slowlog.threshold.query.warn": "5s", "index.search.slowlog.threshold.query.info": "2s" }
Testing slow logs
To test that both search request and shard slow logs are publishing successfully, consider starting with very low values to verify that logs appear in CloudWatch, and then increase the thresholds to more useful levels.
If the logs don't appear, check the following:
-
Does the CloudWatch log group exist? Check the CloudWatch console.
-
Does OpenSearch Service have permissions to write to the log group? Check the OpenSearch Service console.
-
Is the OpenSearch Service domain configured to publish to the log group? Check the OpenSearch Service console, use the AWS CLI
describe-domain-config
option, or callDescribeDomainConfig
using one of the SDKs. -
Are the OpenSearch logging thresholds low enough that your requests are exceeding them?
To review your search request slow log thresholds for a domain, use the following command:
GET
domain-endpoint
/_cluster/settings?flat_settingsTo review your shard slow log thresholds for an index, use the following command:
GET
domain-endpoint
/index
/_settings?pretty
If you want to disable slow logs for an index, return any thresholds that you changed to
their default values of -1
.
Disabling publishing to CloudWatch using the OpenSearch Service console or AWS CLI does not stop OpenSearch from generating logs; it only stops the publishing of those logs. Be sure to check your index settings if you no longer need the shard slow logs, and your domain settings if you no longer need the search request slow logs.
Viewing logs
Viewing the application and slow logs in CloudWatch is just like viewing any other CloudWatch log. For more information, see View Log Data in the Amazon CloudWatch Logs User Guide.
Here are some considerations for viewing the logs:
-
OpenSearch Service publishes only the first 255,000 characters of each line to CloudWatch. Any remaining content is truncated. For audit logs, it's 10,000 characters per message.
-
In CloudWatch, the log stream names have suffixes of
-index-slow-logs
,-search-slow-logs
,-application-logs
, and-audit-logs
to help identify their contents.