Chapter 13. Performance and reliability tuning
13.1. Flow control mechanisms
If logs are produced faster than they can be collected, it can be difficult to predict or control the volume of logs being sent to an output. Not being able to predict or control the volume of logs being sent to an output can result in logs being lost. If there is a system outage and log buffers are accumulated without user control, this can also cause long recovery times and high latency when the connection is restored.
As an administrator, you can limit logging rates by configuring flow control mechanisms for your logging.
13.1.1. Benefits of flow control mechanisms
- The cost and volume of logging can be predicted more accurately in advance.
- Noisy containers cannot produce unbounded log traffic that drowns out other containers.
- Ignoring low-value logs reduces the load on the logging infrastructure.
- High-value logs can be preferred over low-value logs by assigning higher rate limits.
13.1.2. Configuring rate limits
Rate limits are configured per collector, which means that the maximum rate of log collection is the number of collector instances multiplied by the rate limit.
Because logs are collected from each node’s file system, a collector is deployed on each cluster node. For example, in a 3-node cluster, with a maximum rate limit of 10 records per second per collector, the maximum rate of log collection is 30 records per second.
Because the exact byte size of a record as written to an output can vary due to transformations, different encodings, or other factors, rate limits are set in number of records instead of bytes.
You can configure rate limits in the ClusterLogForwarder
custom resource (CR) in two ways:
- Output rate limit
- Limit the rate of outbound logs to selected outputs, for example, to match the network or storage capacity of an output. The output rate limit controls the aggregated per-output rate.
- Input rate limit
- Limit the per-container rate of log collection for selected containers.
13.1.3. Configuring log forwarder output rate limits
You can limit the rate of outbound logs to a specified output by configuring the ClusterLogForwarder
custom resource (CR).
Prerequisites
- You have installed the Red Hat OpenShift Logging Operator.
- You have administrator permissions.
Procedure
Add a
maxRecordsPerSecond
limit value to theClusterLogForwarder
CR for a specified output.The following example shows how to configure a per collector output rate limit for a Kafka broker output named
kafka-example
:Example
ClusterLogForwarder
CRapiVersion: logging.openshift.io/v1 kind: ClusterLogForwarder metadata: # ... spec: # ... outputs: - name: kafka-example 1 type: kafka 2 limit: maxRecordsPerSecond: 1000000 3 # ...
- 1
- The output name.
- 2
- The type of output.
- 3
- The log output rate limit. This value sets the maximum Quantity of logs that can be sent to the Kafka broker per second. This value is not set by default. The default behavior is best effort, and records are dropped if the log forwarder cannot keep up. If this value is
0
, no logs are forwarded.
Apply the
ClusterLogForwarder
CR:Example command
$ oc apply -f <filename>.yaml
Additional resources
13.1.4. Configuring log forwarder input rate limits
You can limit the rate of incoming logs that are collected by configuring the ClusterLogForwarder
custom resource (CR). You can set input limits on a per-container or per-namespace basis.
Prerequisites
- You have installed the Red Hat OpenShift Logging Operator.
- You have administrator permissions.
Procedure
Add a
maxRecordsPerSecond
limit value to theClusterLogForwarder
CR for a specified input.The following examples show how to configure input rate limits for different scenarios:
Example
ClusterLogForwarder
CR that sets a per-container limit for containers with certain labelsapiVersion: logging.openshift.io/v1 kind: ClusterLogForwarder metadata: # ... spec: # ... inputs: - name: <input_name> 1 application: selector: matchLabels: { example: label } 2 containerLimit: maxRecordsPerSecond: 0 3 # ...
- 1
- The input name.
- 2
- A list of labels. If these labels match labels that are applied to a pod, the per-container limit specified in the
maxRecordsPerSecond
field is applied to those containers. - 3
- Configures the rate limit. Setting the
maxRecordsPerSecond
field to0
means that no logs are collected for the container. Setting themaxRecordsPerSecond
field to some other value means that a maximum of that number of records per second are collected for the container.
Example
ClusterLogForwarder
CR that sets a per-container limit for containers in selected namespacesapiVersion: logging.openshift.io/v1 kind: ClusterLogForwarder metadata: # ... spec: # ... inputs: - name: <input_name> 1 application: namespaces: [ example-ns-1, example-ns-2 ] 2 containerLimit: maxRecordsPerSecond: 10 3 - name: <input_name> application: namespaces: [ test ] containerLimit: maxRecordsPerSecond: 1000 # ...
- 1
- The input name.
- 2
- A list of namespaces. The per-container limit specified in the
maxRecordsPerSecond
field is applied to all containers in the namespaces listed. - 3
- Configures the rate limit. Setting the
maxRecordsPerSecond
field to10
means that a maximum of 10 records per second are collected for each container in the namespaces listed.
Apply the
ClusterLogForwarder
CR:Example command
$ oc apply -f <filename>.yaml
13.2. Filtering logs by content
Collecting all logs from a cluster might produce a large amount of data, which can be expensive to transport and store.
You can reduce the volume of your log data by filtering out low priority data that does not need to be stored. Logging provides content filters that you can use to reduce the volume of log data.
Content filters are distinct from input
selectors. input
selectors select or ignore entire log streams based on source metadata. Content filters edit log streams to remove and modify records based on the record content.
Log data volume can be reduced by using one of the following methods:
13.2.1. Configuring content filters to drop unwanted log records
When the drop
filter is configured, the log collector evaluates log streams according to the filters before forwarding. The collector drops unwanted log records that match the specified configuration.
Prerequisites
- You have installed the Red Hat OpenShift Logging Operator.
- You have administrator permissions.
-
You have created a
ClusterLogForwarder
custom resource (CR).
Procedure
Add a configuration for a filter to the
filters
spec in theClusterLogForwarder
CR.The following example shows how to configure the
ClusterLogForwarder
CR to drop log records based on regular expressions:Example
ClusterLogForwarder
CRapiVersion: logging.openshift.io/v1 kind: ClusterLogForwarder metadata: # ... spec: filters: - name: <filter_name> type: drop 1 drop: 2 - test: 3 - field: .kubernetes.labels."foo-bar/baz" 4 matches: .+ 5 - field: .kubernetes.pod_name notMatches: "my-pod" 6 pipelines: - name: <pipeline_name> 7 filterRefs: ["<filter_name>"] # ...
- 1
- Specifies the type of filter. The
drop
filter drops log records that match the filter configuration. - 2
- Specifies configuration options for applying the
drop
filter. - 3
- Specifies the configuration for tests that are used to evaluate whether a log record is dropped.
- If all the conditions specified for a test are true, the test passes and the log record is dropped.
-
When multiple tests are specified for the
drop
filter configuration, if any of the tests pass, the record is dropped. - If there is an error evaluating a condition, for example, the field is missing from the log record being evaluated, that condition evaluates to false.
- 4
- Specifies a dot-delimited field path, which is a path to a field in the log record. The path can contain alpha-numeric characters and underscores (
a-zA-Z0-9_
), for example,.kubernetes.namespace_name
. If segments contain characters outside of this range, the segment must be in quotes, for example,.kubernetes.labels."foo.bar-bar/baz"
. You can include multiple field paths in a singletest
configuration, but they must all evaluate to true for the test to pass and thedrop
filter to be applied. - 5
- Specifies a regular expression. If log records match this regular expression, they are dropped. You can set either the
matches
ornotMatches
condition for a singlefield
path, but not both. - 6
- Specifies a regular expression. If log records do not match this regular expression, they are dropped. You can set either the
matches
ornotMatches
condition for a singlefield
path, but not both. - 7
- Specifies the pipeline that the
drop
filter is applied to.
Apply the
ClusterLogForwarder
CR by running the following command:$ oc apply -f <filename>.yaml
Additional examples
The following additional example shows how you can configure the drop
filter to only keep higher priority log records:
apiVersion: logging.openshift.io/v1 kind: ClusterLogForwarder metadata: # ... spec: filters: - name: important type: drop drop: test: - field: .message notMatches: "(?i)critical|error" - field: .level matches: "info|warning" # ...
In addition to including multiple field paths in a single test
configuration, you can also include additional tests that are treated as OR checks. In the following example, records are dropped if either test
configuration evaluates to true. However, for the second test
configuration, both field specs must be true for it to be evaluated to true:
apiVersion: logging.openshift.io/v1 kind: ClusterLogForwarder metadata: # ... spec: filters: - name: important type: drop drop: test: - field: .kubernetes.namespace_name matches: "^open" test: - field: .log_type matches: "application" - field: .kubernetes.pod_name notMatches: "my-pod" # ...
13.2.2. Configuring content filters to prune log records
When the prune
filter is configured, the log collector evaluates log streams according to the filters before forwarding. The collector prunes log records by removing low value fields such as pod annotations.
Prerequisites
- You have installed the Red Hat OpenShift Logging Operator.
- You have administrator permissions.
-
You have created a
ClusterLogForwarder
custom resource (CR).
Procedure
Add a configuration for a filter to the
prune
spec in theClusterLogForwarder
CR.The following example shows how to configure the
ClusterLogForwarder
CR to prune log records based on field paths:ImportantIf both are specified, records are pruned based on the
notIn
array first, which takes precedence over thein
array. After records have been pruned by using thenotIn
array, they are then pruned by using thein
array.Example
ClusterLogForwarder
CRapiVersion: logging.openshift.io/v1 kind: ClusterLogForwarder metadata: # ... spec: filters: - name: <filter_name> type: prune 1 prune: 2 in: [.kubernetes.annotations, .kubernetes.namespace_id] 3 notIn: [.kubernetes,.log_type,.message,."@timestamp"] 4 pipelines: - name: <pipeline_name> 5 filterRefs: ["<filter_name>"] # ...
- 1
- Specify the type of filter. The
prune
filter prunes log records by configured fields. - 2
- Specify configuration options for applying the
prune
filter. Thein
andnotIn
fields are specified as arrays of dot-delimited field paths, which are paths to fields in log records. These paths can contain alpha-numeric characters and underscores (a-zA-Z0-9_
), for example,.kubernetes.namespace_name
. If segments contain characters outside of this range, the segment must be in quotes, for example,.kubernetes.labels."foo.bar-bar/baz"
. - 3
- Optional: Any fields that are specified in this array are removed from the log record.
- 4
- Optional: Any fields that are not specified in this array are removed from the log record.
- 5
- Specify the pipeline that the
prune
filter is applied to.
Apply the
ClusterLogForwarder
CR by running the following command:$ oc apply -f <filename>.yaml
13.2.3. Additional resources
13.3. Filtering logs by metadata
You can filter logs in the ClusterLogForwarder
CR to select or ignore an entire log stream based on the metadata by using the input
selector. As an administrator or developer, you can include or exclude the log collection to reduce the memory and CPU load on the collector.
You can use this feature only if the Vector collector is set up in your logging deployment.
input
spec filtering is different from content filtering. input
selectors select or ignore entire log streams based on the source metadata. Content filters edit the log streams to remove and modify the records based on the record content.
13.3.1. Filtering application logs at input by including or excluding the namespace or container name
You can include or exclude the application logs based on the namespace and container name by using the input
selector.
Prerequisites
- You have installed the Red Hat OpenShift Logging Operator.
- You have administrator permissions.
-
You have created a
ClusterLogForwarder
custom resource (CR).
Procedure
Add a configuration to include or exclude the namespace and container names in the
ClusterLogForwarder
CR.The following example shows how to configure the
ClusterLogForwarder
CR to include or exclude namespaces and container names:Example
ClusterLogForwarder
CRapiVersion: "logging.openshift.io/v1" kind: ClusterLogForwarder # ... spec: inputs: - name: mylogs application: includes: - namespace: "my-project" 1 container: "my-container" 2 excludes: - container: "other-container*" 3 namespace: "other-namespace" 4 # ...
Apply the
ClusterLogForwarder
CR by running the following command:$ oc apply -f <filename>.yaml
The excludes
option takes precedence over includes
.
13.3.2. Filtering application logs at input by including either the label expressions or matching label key and values
You can include the application logs based on the label expressions or a matching label key and its values by using the input
selector.
Prerequisites
- You have installed the Red Hat OpenShift Logging Operator.
- You have administrator permissions.
-
You have created a
ClusterLogForwarder
custom resource (CR).
Procedure
Add a configuration for a filter to the
input
spec in theClusterLogForwarder
CR.The following example shows how to configure the
ClusterLogForwarder
CR to include logs based on label expressions or matched label key/values:Example
ClusterLogForwarder
CRapiVersion: "logging.openshift.io/v1" kind: ClusterLogForwarder # ... spec: inputs: - name: mylogs application: selector: matchExpressions: - key: env 1 operator: In 2 values: [“prod”, “qa”] 3 - key: zone operator: NotIn values: [“east”, “west”] matchLabels: 4 app: one name: app1 # ...
Apply the
ClusterLogForwarder
CR by running the following command:$ oc apply -f <filename>.yaml
13.3.3. Filtering the audit and infrastructure log inputs by source
You can define the list of audit
and infrastructure
sources to collect the logs by using the input
selector.
Prerequisites
- You have installed the Red Hat OpenShift Logging Operator.
- You have administrator permissions.
-
You have created a
ClusterLogForwarder
custom resource (CR).
Procedure
Add a configuration to define the
audit
andinfrastructure
sources in theClusterLogForwarder
CR.The following example shows how to configure the
ClusterLogForwarder
CR to defineaduit
andinfrastructure
sources:Example
ClusterLogForwarder
CRapiVersion: "logging.openshift.io/v1" kind: ClusterLogForwarder # ... spec: inputs: - name: mylogs1 infrastructure: sources: 1 - node - name: mylogs2 audit: sources: 2 - kubeAPI - openshiftAPI - ovn # ...
- 1
- Specifies the list of infrastructure sources to collect. The valid sources include:
-
node
: Journal log from the node -
container
: Logs from the workloads deployed in the namespaces
-
- 2
- Specifies the list of audit sources to collect. The valid sources include:
-
kubeAPI
: Logs from the Kubernetes API servers -
openshiftAPI
: Logs from the OpenShift API servers -
auditd
: Logs from a node auditd service -
ovn
: Logs from an open virtual network service
-
Apply the
ClusterLogForwarder
CR by running the following command:$ oc apply -f <filename>.yaml