The agent is configured primarily by a YAML document conventionally located at
/etc/signalfx/agent.yaml. The location of the config file can be specified
by the -config flag to the agent binary (signalfx-agent).
By default, the Smart Agent will send data to the us0 realm.
If you are not in this realm, you will need to explicitly set the
signalFxRealm option in your config like this:
signalFxRealm: YOUR_SIGNALFX_REALM
To determine if you are in a different realm and need to explicitly set the endpoints, check your profile page in the SignalFx web application.
If you want to explicitly set the ingest, API server, and trace endpoint URLs, you can set them individually like so:
ingestUrl: "https://ingest.YOUR_SIGNALFX_REALM.signalfx.com"
apiUrl: "https://api.YOUR_SIGNALFX_REALM.signalfx.com"
traceEndpointUrl: "https://ingest.YOUR_SIGNALFX_REALM.signalfx.com/v1/trace"
They will default to the endpoints for the realm configured in signalFxRealm
if not set.
| Config option | Required | Type | Description |
|---|---|---|---|
signalFxAccessToken |
no | string | The access token for the org that should receive the metrics emitted by the agent. |
ingestUrl |
no | string | The URL of SignalFx ingest server. Should be overridden if using the SignalFx Gateway. If not set, this will be determined by the signalFxRealm option below. If you want to send trace spans to a different location, set the traceEndpointUrl option. If you want to send events to a different location, set the eventEndpointUrl option. |
eventEndpointUrl |
no | string | The full URL (including path) to the event ingest server. If this is not set, all events will be sent to the same place as ingestUrl above. |
traceEndpointUrl |
no | string | The full URL (including path) to the trace ingest server. If this is not set, all trace spans will be sent to the same place as ingestUrl above. |
apiUrl |
no | string | The SignalFx API base URL. If not set, this will determined by the signalFxRealm option below. |
signalFxRealm |
no | string | The SignalFx Realm that the organization you want to send to is a part of. This defaults to the original realm (us0) but if you are setting up the agent for the first time, you quite likely need to change this. (default: "us0") |
hostname |
no | string | The hostname that will be reported as the host dimension. If blank, this will be auto-determined by the agent based on a reverse lookup of the machine's IP address. |
useFullyQualifiedHost |
no | bool | If true (the default), and the hostname option is not set, the hostname will be determined by doing a reverse DNS query on the IP address that is returned by querying for the bare hostname. This is useful in cases where the hostname reported by the kernel is a short name. (default: true) |
disableHostDimensions |
no | bool | Our standard agent model is to collect metrics for services running on the same host as the agent. Therefore, host-specific dimensions (e.g. host, AWSUniqueId, etc) are automatically added to every datapoint that is emitted from the agent by default. Set this to true if you are using the agent primarily to monitor things on other hosts. You can set this option at the monitor level as well. (default: false) |
intervalSeconds |
no | integer | How often to send metrics to SignalFx. Monitors can override this individually. (default: 10) |
cloudMetadataTimeout |
no | int64 | This flag sets the HTTP timeout duration for metadata queries from AWS, Azure and GCP. This should be a duration string that is accepted by https://golang.org/pkg/time/#ParseDuration (default: "2s") |
globalDimensions |
no | map of strings | Dimensions (key:value pairs) that will be added to every datapoint emitted by the agent. To specify that all metrics should be high-resolution, add the dimension sf_hires: 1 |
globalSpanTags |
no | map of strings | Tags (key:value pairs) that will be added to every span emitted by the agent. |
cluster |
no | string | The logical environment/cluster that this agent instance is running in. All of the services that this instance monitors should be in the same environment as well. This value, if provided, will be synced as a property onto the host dimension, or onto any cloud-provided specific dimensions (AWSUniqueId, gcp_id, and azure_resource_id) when available. Example values: "prod-usa", "dev" |
syncClusterOnHostDimension |
no | bool | If true, force syncing of the cluster property on the host dimension, even when cloud-specific dimensions are present. (default: false) |
validateDiscoveryRules |
no | bool | If true, a warning will be emitted if a discovery rule contains variables that will never possibly match a rule. If using multiple observers, it is convenient to set this to false to suppress spurious errors. (default: false) |
observers |
no | list of objects (see below) | A list of observers to use (see observer config) |
monitors |
no | list of objects (see below) | A list of monitors to use (see monitor config) |
writer |
no | object (see below) | Configuration of the datapoint/event writer |
logging |
no | object (see below) | Log configuration |
collectd |
no | object (see below) | Configuration of the managed collectd subprocess |
enableBuiltInFiltering |
no | bool | This must be unset or explicitly set to true. In prior versions of the agent, there was a filtering mechanism that relied heavily on an external whitelist.json file to determine which metrics were sent by default. This is all inherent to the agent now and the old style of filtering is no longer available. (default: true) |
metricsToInclude |
no | list of objects (see below) | A list of metric filters that will include metrics. These filters take priority over the filters specified in metricsToExclude. |
metricsToExclude |
no | list of objects (see below) | A list of metric filters |
propertiesToExclude |
no | list of objects (see below) | A list of properties filters |
internalStatusHost |
no | string | The host on which the internal status server will listen. The internal status HTTP server serves internal metrics and diagnostic information about the agent and can be scraped by the internal-metrics monitor. Can be set to 0.0.0.0 if you want to monitor the agent from another host. If you set this to blank/null, the internal status server will not be started. See internalStatusPort. (default: "localhost") |
internalStatusPort |
no | integer | The port on which the internal status server will listen. See internalStatusHost. (default: 8095) |
profiling |
no | bool | Enables Go pprof endpoint on port 6060 that serves profiling data for development (default: false) |
profilingHost |
no | string | The host/ip address for the pprof profile server to listen on. profiling must be enabled for this to have any effect. (default: "127.0.0.1") |
profilingPort |
no | integer | The port for the pprof profile server to listen on. profiling must be enabled for this to have any effect. (default: 6060) |
bundleDir |
no | string | Path to the directory holding the agent dependencies. This will normally be derived automatically. Overrides the envvar SIGNALFX_BUNDLE_DIR if set. |
scratch |
no | any | This exists purely to give the user a place to put common yaml values to reference in other parts of the config file. |
configSources |
no | object (see below) | Configuration of remote config stores |
procPath |
no | string | Path to the host's /proc filesystem. This is useful for containerized environments. (default: "/proc") |
etcPath |
no | string | Path to the host's /etc directory. This is useful for containerized environments. (default: "/etc") |
varPath |
no | string | Path to the host's /var directory. This is useful for containerized environments. (default: "/var") |
runPath |
no | string | Path to the host's /run directory. This is useful for containerized environments. (default: "/run") |
sysPath |
no | string | Path to the host's /sys directory. This is useful for containerized environments. (default: "/sys") |
The nested observers config object has the following fields:
The following are generic options that apply to all observers. Each observer type has its own set of additional configuration options, detailed in Observer Config.
| Config option | Required | Type | Description |
|---|---|---|---|
type |
no | string | The type of the observer |
The nested monitors config object has the following fields:
The following are generic options that apply to all monitors. Each monitor type has its own set of additional configuration options, detailed in Monitor Config.
| Config option | Required | Type | Description |
|---|---|---|---|
type |
no | string | The type of the monitor |
discoveryRule |
no | string | The rule used to match up this configuration with a discovered endpoint. If blank, the configuration will be run immediately when the agent is started. If multiple endpoints match this rule, multiple instances of the monitor type will be created with the same configuration (except different host/port). |
validateDiscoveryRule |
no | bool | If true, a warning will be emitted if a discovery rule contains variables that will never possibly match a rule. If using multiple observers, it is convenient to set this to false to suppress spurious errors. The top-level setting validateDiscoveryRules acts as a default if this isn't set. (default: "false") |
extraDimensions |
no | map of strings | A set of extra dimensions (key:value pairs) to include on datapoints emitted by the monitor(s) created from this configuration. To specify metrics from this monitor should be high-resolution, add the dimension sf_hires: 1 |
extraSpanTags |
no | map of strings | A set of extra span tags (key:value pairs) to include on spans emitted by the monitor(s) created from this configuration. |
extraSpanTagsFromEndpoint |
no | map of strings | A mapping of extra span tag names to a discovery rule expression that is used to derive the value of the span tag. For example, to use a certain container label as a span tag, you could use something like this in your monitor config block: extraSpanTagsFromEndpoint: {env: 'Get(container_labels, "myapp.com/environment")'}. This only applies when the monitor has a discoveryRule or was dynamically instantiated by an endpoint. It does nothing, for example, in the signalfx-forwarder montior. |
defaultSpanTags |
no | map of strings | A set of default span tags (key:value pairs) to include on spans emitted by the monitor(s) created from this configuration. |
defaultSpanTagsFromEndpoint |
no | map of strings | A mapping of default span tag names to a discovery rule expression that is used to derive the default value of the span tag. For example, to use a certain container label as a span tag, you could use something like this in your monitor config block: defaultSpanTagsFromEndpoint: {env: 'Get(container_labels, "myapp.com/environment")'} This only applies when the monitor has a discoveryRule or was dynamically instantiated by an endpoint. It does nothing, for example, in the signalfx-forwarder montior. |
extraDimensionsFromEndpoint |
no | map of strings | A mapping of extra dimension names to a discovery rule expression that is used to derive the value of the dimension. For example, to use a certain container label as a dimension, you could use something like this in your monitor config block: extraDimensionsFromEndpoint: {env: 'Get(container_labels, "myapp.com/environment")'}. This only applies when the monitor has a discoveryRule or was dynamically instantiated by an endpoint. It does nothing, for example, in the signalfx-forwarder montior. |
configEndpointMappings |
no | map of strings | A set of mappings from a configuration option on this monitor to attributes of a discovered endpoint. The keys are the config option on this monitor and the value can be any valid expression used in discovery rules. |
intervalSeconds |
no | integer | The interval (in seconds) at which to emit datapoints from the monitor(s) created by this configuration. If not set (or set to 0), the global agent intervalSeconds config option will be used instead. (default: 0) |
solo |
no | bool | If one or more configurations have this set to true, only those configurations will be considered. This setting can be useful for testing. (default: false) |
datapointsToExclude |
no | list of objects (see below) | A list of datapoint filters. These filters allow you to comprehensively define which datapoints to exclude by metric name or dimension set, as well as the ability to define overrides to re-include metrics excluded by previous patterns within the same filter item. See monitor filtering for examples and more information. |
disableHostDimensions |
no | bool | Some monitors pull metrics from services not running on the same host and should not get the host-specific dimensions set on them (e.g. host, AWSUniqueId, etc). Setting this to true causes those dimensions to be omitted. You can disable this globally with the disableHostDimensions option on the top level of the config. (default: false) |
disableEndpointDimensions |
no | bool | This can be set to true if you don't want to include the dimensions that are specific to the endpoint that was discovered by an observer. This is useful when you have an endpoint whose identity is not particularly important since it acts largely as a proxy or adapter for other metrics. (default: false) |
metricNameTransformations |
no | map | A map from original metric name to a replacement value. The keys are intepreted as regular expressions and the values can contain backreferences. This means that you should escape any RE characters in the original metric name with \ (the most common escape necessary will be \. as period is interpreted as "all characters" if unescaped). The Go regexp language, and backreferences are of the form $1. If there are multiple entries in list of maps, they will each be run in sequence, using the transformation from the previous entry as the input the subsequent transformation. To add a common prefix to all metrics coming out of a monitor, use a mapping like this: (.*): myprefix.$1 |
dimensionTransformations |
no | map of strings | A map from dimension names emitted by the monitor to the desired dimension name that will be emitted in the datapoint that goes to SignalFx. This can be useful if you have custom metrics from your applications and want to make the dimensions from a monitor match those. Also can be useful when scraping free-form metrics, say with the prometheus-exporter monitor. Right now, only static key/value transformations are supported. Note that filtering by dimensions will be done on the original dimension name and not the new name. Note that it is possible to remove unwanted dimensions via this configuration, by making the desired dimension name an empty string. |
extraMetrics |
no | list of strings | Extra metrics to enable besides the default included ones. This is an overridable filter. |
extraGroups |
no | list of strings | Extra metric groups to enable in addition to the metrics that are emitted by default. A metric group is simply a collection of metrics, and they are defined in each monitor's documentation. |
The nested datapointsToExclude config object has the following fields:
| Config option | Required | Type | Description |
|---|---|---|---|
dimensions |
no | map of any | A map of dimension key/values to match against. All key/values must match a datapoint for it to be matched. The map values can be either a single string or a list of strings. |
metricNames |
no | list of strings | A list of metric names to match against |
metricName |
no | string | A single metric name to match against |
monitorType |
no | string | (Only applicable for the top level filters) Limits this scope of the filter to datapoints from a specific monitor. If specified, any datapoints not from this monitor type will never match against this filter. |
negated |
no | bool | (Only applicable for the top level filters) Negates the result of the match so that it matches all datapoints that do NOT match the metric name and dimension values given. This does not negate monitorType, if given. (default: false) |
The nested writer config object has the following fields:
| Config option | Required | Type | Description |
|---|---|---|---|
datapointMaxBatchSize |
no | integer | The maximum number of datapoints to include in a batch before sending the batch to the ingest server. Smaller batch sizes than this will be sent if datapoints originate in smaller chunks. (default: 1000) |
maxDatapointsBuffered |
no | integer | The maximum number of datapoints that are allowed to be buffered in the agent (i.e. received from a monitor but have not yet received confirmation of successful receipt by the target ingest/gateway server downstream). Any datapoints that come in beyond this number will overwrite existing datapoints if they have not been sent yet, starting with the oldest. (default: 25000) |
traceSpanMaxBatchSize |
no | integer | The analogue of datapointMaxBatchSize for trace spans. (default: 1000) |
traceExportFormat |
no | string | Format to export traces in. Choices are "zipkin" and "sapm" (default: "zipkin") |
datapointMaxRequests |
no | integer | Deprecated: use maxRequests instead. (default: 0) |
maxRequests |
no | integer | The maximum number of concurrent requests to make to a single ingest server with datapoints/events/trace spans. This number multiplied by datapointMaxBatchSize is more or less the maximum number of datapoints that can be "in-flight" at any given time. Same thing for the traceSpanMaxBatchSize option and trace spans. (default: 10) |
timeout |
no | int64 | Timeout specifies a time limit for requests made to the ingest server. The timeout includes connection time, any redirects, and reading the response body. Default is 5 seconds, a Timeout of zero means no timeout. (default: "5s") |
eventSendIntervalSeconds |
no | integer | The agent does not send events immediately upon a monitor generating them, but buffers them and sends them in batches. The lower this number, the less delay for events to appear in SignalFx. (default: 1) |
propertiesMaxRequests |
no | unsigned integer | The analogue of maxRequests for dimension property requests. (default: 20) |
propertiesMaxBuffered |
no | unsigned integer | How many dimension property updates to hold pending being sent before dropping subsequent property updates. Property updates will be resent eventually and they are slow to change so dropping them (esp on agent start up) usually isn't a big deal. (default: 10000) |
propertiesSendDelaySeconds |
no | unsigned integer | How long to wait for property updates to be sent once they are generated. Any duplicate updates to the same dimension within this time frame will result in the latest property set being sent. This helps prevent spurious updates that get immediately overwritten by very flappy property generation. (default: 30) |
propertiesHistorySize |
no | unsigned integer | Properties that are synced to SignalFx are cached to prevent duplicate requests from being sent, causing unnecessary load on our backend. (default: 10000) |
logDatapoints |
no | bool | If the log level is set to debug and this is true, all datapoints generated by the agent will be logged. (default: false) |
logEvents |
no | bool | The analogue of logDatapoints for events. (default: false) |
logTraceSpans |
no | bool | The analogue of logDatapoints for trace spans. (default: false) |
logDimensionUpdates |
no | bool | If true, dimension updates will be logged at the INFO level. (default: false) |
logDroppedDatapoints |
no | bool | If true, and the log level is debug, filtered out datapoints will be logged. (default: false) |
addGlobalDimensionsAsSpanTags |
no | bool | If true, the dimensions specified in the top-level globalDimensions configuration will be added to the tag set of all spans that are emitted by the writer. If this is false, only the "host id" dimensions such as host, AwsUniqueId, etc. are added to the span tags. (default: false) |
sendTraceHostCorrelationMetrics |
no | bool | Whether to send host correlation metrics to correlate traced services with the underlying host (default: true) |
staleServiceTimeout |
no | int64 | How long to wait after a trace span's service name is last seen to continue sending the correlation datapoints for that service. This should be a duration string that is accepted by https://golang.org/pkg/time/#ParseDuration. This option is irrelevant if sendTraceHostCorrelationMetrics is false. (default: "5m") |
traceHostCorrelationPurgeInterval |
no | int64 | How frequently to purge host correlation caches that are generated from the service and environment names seen in trace spans sent through or by the agent. This should be a duration string that is accepted by https://golang.org/pkg/time/#ParseDuration. (default: "1m") |
traceHostCorrelationMetricsInterval |
no | int64 | How frequently to send host correlation metrics that are generated from the service name seen in trace spans sent through or by the agent. This should be a duration string that is accepted by https://golang.org/pkg/time/#ParseDuration. This option is irrelevant if sendTraceHostCorrelationMetrics is false. (default: "1m") |
traceHostCorrelationMaxRequestRetries |
no | unsigned integer | How many times to retry requests related to trace host correlation (default: 2) |
maxTraceSpansInFlight |
no | unsigned integer | How many trace spans are allowed to be in the process of sending. While this number is exceeded, the oldest spans will be discarded to accommodate new spans generated to avoid memory exhaustion. If you see log messages about "Aborting pending trace requests..." or "Dropping new trace spans..." it means that the downstream target for traces is not able to accept them fast enough. Usually if the downstream is offline you will get connection refused errors and most likely spans will not build up in the agent (there is no retry mechanism). In the case of slow downstreams, you might be able to increase maxRequests to increase the concurrent stream of spans downstream (if the target can make efficient use of additional connections) or, less likely, increase traceSpanMaxBatchSize if your batches are maxing out (turn on debug logging to see the batch sizes being sent) and being split up too much. If neither of those options helps, your downstream is likely too slow to handle the volume of trace spans and should be upgraded to more powerful hardware/networking. (default: 100000) |
splunk |
no | object (see below) | Configures the writer specifically writing to Splunk. |
signalFxEnabled |
no | bool | If set to false, output to SignalFx will be disabled. (default: true) |
The nested splunk config object has the following fields:
| Config option | Required | Type | Description |
|---|---|---|---|
enabled |
no | bool | Enable logging to a Splunk Enterprise instance (default: false) |
url |
no | string | Full URL (including path) of Splunk HTTP Event Collector (HEC) endpoint |
token |
no | string | Splunk HTTP Event Collector token |
source |
no | string | Splunk source field value, description of the source of the event |
sourceType |
no | string | Splunk source type, optional name of a sourcetype field value |
index |
no | string | Splunk index, optional name of the Splunk index to store the event in |
eventsIndex |
no | string | Splunk index, specifically for traces (must be event type) |
eventsSource |
no | string | Splunk source field value, description of the source of the trace |
eventsSourceType |
no | string | Splunk trace source type, optional name of a sourcetype field value |
skipTLSVerify |
no | bool | Skip verifying the certificate of the HTTP Event Collector (default: false) |
maxBuffered |
no | integer | The maximum number of Splunk log entries of all types (e.g. metric, event) to be buffered before old events are dropped. Defaults to the writer.maxDatapointsBuffered config if not specified. (default: 0) |
maxRequests |
no | integer | The maximum number of simultaneous requests to the Splunk HEC endpoint. Defaults to the writer.maxBuffered config if not specified. (default: 0) |
maxBatchSize |
no | integer | The maximum number of Splunk log entries to submit in one request to the HEC (default: 0) |
The nested logging config object has the following fields:
| Config option | Required | Type | Description |
|---|---|---|---|
level |
no | string | Valid levels include debug, info, warn, error. Note that debug logging may leak sensitive configuration (e.g. passwords) to the agent output. (default: "info") |
format |
no | string | The log output format to use. Valid values are: text, json. (default: "text") |
The nested collectd config object has the following fields:
| Config option | Required | Type | Description |
|---|---|---|---|
disableCollectd |
no | bool | If you won't be using any collectd monitors, this can be set to true to prevent collectd from pre-initializing (default: false) |
timeout |
no | integer | How many read intervals before abandoning a metric. Doesn't affect much in normal usage. See Timeout. (default: 40) |
readThreads |
no | integer | Number of threads dedicated to executing read callbacks. See ReadThreads (default: 5) |
writeThreads |
no | integer | Number of threads dedicated to writing value lists to write callbacks. This should be much less than readThreads because writing is batched in the write_http plugin that writes back to the agent. See WriteThreads. (default: 2) |
writeQueueLimitHigh |
no | integer | The maximum numbers of values in the queue to be written back to the agent from collectd. Since the values are written to a local socket that the agent exposes, there should be almost no queuing and the default should be more than sufficient. See WriteQueueLimitHigh (default: 500000) |
writeQueueLimitLow |
no | integer | The lowest number of values in the collectd queue before which metrics begin being randomly dropped. See WriteQueueLimitLow (default: 400000) |
logLevel |
no | string | Collectd's log level -- info, notice, warning, or err (default: "notice") |
intervalSeconds |
no | integer | A default read interval for collectd plugins. If zero or undefined, will default to the global agent interval. Some collectd python monitors do not support overridding the interval at the monitor level, but this setting will apply to them. (default: 0) |
writeServerIPAddr |
no | string | The local IP address of the server that the agent exposes to which collectd will send metrics. This defaults to an arbitrary address in the localhost subnet, but can be overridden if needed. (default: "127.9.8.7") |
writeServerPort |
no | integer | The port of the agent's collectd metric sink server. If set to zero (the default) it will allow the OS to assign it a free port. (default: 0) |
configDir |
no | string | This is where the agent will write the collectd config files that it manages. If you have secrets in those files, consider setting this to a path on a tmpfs mount. The files in this directory should be considered transient -- there is no value in editing them by hand. If you want to add your own collectd config, see the collectd/custom monitor. (default: "/var/run/signalfx-agent/collectd") |
The nested metricsToInclude config object has the following fields:
| Config option | Required | Type | Description |
|---|---|---|---|
dimensions |
no | map of any | A map of dimension key/values to match against. All key/values must match a datapoint for it to be matched. The map values can be either a single string or a list of strings. |
metricNames |
no | list of strings | A list of metric names to match against |
metricName |
no | string | A single metric name to match against |
monitorType |
no | string | (Only applicable for the top level filters) Limits this scope of the filter to datapoints from a specific monitor. If specified, any datapoints not from this monitor type will never match against this filter. |
negated |
no | bool | (Only applicable for the top level filters) Negates the result of the match so that it matches all datapoints that do NOT match the metric name and dimension values given. This does not negate monitorType, if given. (default: false) |
The nested metricsToExclude config object has the following fields:
For more information on filtering see Datapoint Filtering.
| Config option | Required | Type | Description |
|---|---|---|---|
dimensions |
no | map of any | A map of dimension key/values to match against. All key/values must match a datapoint for it to be matched. The map values can be either a single string or a list of strings. |
metricNames |
no | list of strings | A list of metric names to match against |
metricName |
no | string | A single metric name to match against |
monitorType |
no | string | (Only applicable for the top level filters) Limits this scope of the filter to datapoints from a specific monitor. If specified, any datapoints not from this monitor type will never match against this filter. |
negated |
no | bool | (Only applicable for the top level filters) Negates the result of the match so that it matches all datapoints that do NOT match the metric name and dimension values given. This does not negate monitorType, if given. (default: false) |
The nested propertiesToExclude config object has the following fields:
| Config option | Required | Type | Description |
|---|---|---|---|
propertyName |
no | string | A single property name to match (default: "*") |
propertyValue |
no | string | A property value to match (default: "*") |
dimensionName |
no | string | A dimension name to match (default: "*") |
dimensionValue |
no | string | A dimension value to match (default: "*") |
The nested configSources config object has the following fields:
For more information about how to use config sources, see Remote Config.
| Config option | Required | Type | Description |
|---|---|---|---|
watch |
no | bool | Whether to watch config sources for changes. If this is true and any of the config changes (either the main agent.yaml, or remote config values), the agent will dynamically reconfigure itself with minimal disruption. This is generally better than restarting the agent on config changes since that can result in larger gaps in metric data. The main disadvantage of watching is slightly greater network and compute resource usage. This option is not itself watched for changes. If you change the value of this option, you must restart the agent. (default: true) |
file |
no | object (see below) | Configuration for other file sources |
zookeeper |
no | object (see below) | Configuration for a Zookeeper remote config source |
etcd2 |
no | object (see below) | Configuration for an Etcd 2 remote config source |
consul |
no | object (see below) | Configuration for a Consul remote config source |
vault |
no | object (see below) | Configuration for a Hashicorp Vault remote config source |
The nested file config object has the following fields:
| Config option | Required | Type | Description |
|---|---|---|---|
pollRateSeconds |
no | integer | How often to poll files (in seconds) to test for changes. There are so many edge cases that break inotify that it is more robust to simply poll files than rely on that. This option is not subject to watching and changes to it will require an agent restart. (default: 5) |
The nested zookeeper config object has the following fields:
| Config option | Required | Type | Description |
|---|---|---|---|
endpoints |
no | list of strings | A list of Zookeeper servers to use for the client |
timeoutSeconds |
no | unsigned integer | Client timeout (default: 10) |
The nested etcd2 config object has the following fields:
| Config option | Required | Type | Description |
|---|---|---|---|
endpoints |
no | list of strings | A list of Etcd2 servers to use |
username |
no | string | An optional username to use when connecting |
password |
no | string | An optional password to use when connecting |
The nested consul config object has the following fields:
| Config option | Required | Type | Description |
|---|---|---|---|
endpoint |
no | string | A Consul server URL |
username |
no | string | An optional username to use when connecting |
password |
no | string | An optional password to use when connecting |
token |
no | string | An authentication token, if needed |
datacenter |
no | string | The Consul datacenter to use |
The nested vault config object has the following fields:
| Config option | Required | Type | Description |
|---|---|---|---|
vaultAddr |
no | string | The Vault Address. Can also be provided by the standard Vault envvar VAULT_ADDR. This option takes priority over the envvar if provided. |
vaultToken |
no | string | The Vault token, can also be provided by it the standard Vault envvar VAULT_TOKEN. This option takes priority over the envvar if provided. |
kvV2PollInterval |
no | int64 | The polling interval for checking KV V2 secrets for a new version. This can be any string value that can be parsed by https://golang.org/pkg/time/#ParseDuration. (default: "60s") |
authMethod |
no | string | The authetication method to use, if any, to obtain the Vault token. If vaultToken is specified above, this option will have no effect. Currently supported values are: iam. |
iam |
no | object (see below) | Further config options for the iam auth method. These options are identical to the CLI helper tool options |
gcp |
no | object (see below) | Further config options for the gcp auth method. These options are identical to the CLI helper tool options. You must provide a valid GCP IAM credential JSON either explicitly via the credentials option (not recommended), or through any GCP Application Default Credentials. |
The nested iam config object has the following fields:
| Config option | Required | Type | Description |
|---|---|---|---|
awsAccessKeyId |
no | string | Explicit AWS access key ID |
awsSecretAccessKey |
no | string | Explicit AWS secret access key |
awsSecurityToken |
no | string | Explicit AWS security token for temporary credentials |
headerValue |
no | string | Value for the x-vault-aws-iam-server-id header in requests |
mount |
no | string | Path where the AWS credential method is mounted. This is usually provided via the -path flag in the "vault login" command, but it can be specified here as well. If specified here, it takes precedence over the value for -path. The default value is "aws". |
role |
no | string | Name of the Vault role to request a token against |
The nested gcp config object has the following fields:
| Config option | Required | Type | Description |
|---|---|---|---|
role |
no | string | Required. The name of the role you're requesting a token for. |
mount |
no | string | This is usually provided via the -path flag in the "vault login" command, but it can be specified here as well. If specified here, it takes precedence over the value for -path. Defaults to gcp. |
credentials |
no | string | Explicitly specified GCP credentials in JSON string format (not recommended) |
jwt_exp |
no | integer | Time until the generated JWT expires in minutes. The given IAM role will have a max_jwt_exp field, the time in minutes that all valid authentication JWTs must expire within (from time of authentication). Defaults to 15 minutes, the default max_jwt_exp for a role. Must be less than an hour. |
service_account |
no | string | Service account to generate a JWT for. Defaults to credentials "client_email" if "credentials" specified and this value is not. The actual credential must have the "iam.serviceAccounts.signJWT" permissions on this service account. |
project |
no | string | Project for the service account who will be authenticating to Vault. Defaults to the credential's "project_id" (if credentials are specified)." |
Here is an autogenerated example of a YAML config file, with default values where applicable:
signalFxAccessToken:
ingestUrl:
eventEndpointUrl:
traceEndpointUrl:
apiUrl:
signalFxRealm: "us0"
hostname:
useFullyQualifiedHost:
disableHostDimensions: false
intervalSeconds: 10
cloudMetadataTimeout: "2s"
globalDimensions:
globalSpanTags:
cluster:
syncClusterOnHostDimension: false
validateDiscoveryRules: false
observers: []
monitors: []
writer:
datapointMaxBatchSize: 1000
maxDatapointsBuffered: 25000
traceSpanMaxBatchSize: 1000
traceExportFormat: "zipkin"
datapointMaxRequests: 0
maxRequests: 10
timeout: "5s"
eventSendIntervalSeconds: 1
propertiesMaxRequests: 20
propertiesMaxBuffered: 10000
propertiesSendDelaySeconds: 30
propertiesHistorySize: 10000
logDatapoints: false
logEvents: false
logTraceSpans: false
logDimensionUpdates: false
logDroppedDatapoints: false
addGlobalDimensionsAsSpanTags: false
sendTraceHostCorrelationMetrics: true
staleServiceTimeout: "5m"
traceHostCorrelationPurgeInterval: "1m"
traceHostCorrelationMetricsInterval: "1m"
traceHostCorrelationMaxRequestRetries: 2
maxTraceSpansInFlight: 100000
splunk:
enabled: false
url:
token:
source:
sourceType:
index:
eventsIndex:
eventsSource:
eventsSourceType:
skipTLSVerify: false
maxBuffered: 0
maxRequests: 0
maxBatchSize: 0
signalFxEnabled: true
logging:
level: "info"
format: "text"
collectd:
disableCollectd: false
timeout: 40
readThreads: 5
writeThreads: 2
writeQueueLimitHigh: 500000
writeQueueLimitLow: 400000
logLevel: "notice"
intervalSeconds: 0
writeServerIPAddr: "127.9.8.7"
writeServerPort: 0
configDir: "/var/run/signalfx-agent/collectd"
enableBuiltInFiltering: true
metricsToInclude: []
metricsToExclude: []
propertiesToExclude: []
internalStatusHost: "localhost"
internalStatusPort: 8095
profiling: false
profilingHost: "127.0.0.1"
profilingPort: 6060
bundleDir:
scratch:
configSources:
watch: true
file:
pollRateSeconds: 5
zookeeper:
endpoints: []
timeoutSeconds: 10
etcd2:
endpoints: []
username:
password:
consul:
endpoint:
username:
password:
token:
datacenter:
vault:
vaultAddr:
vaultToken:
kvV2PollInterval: "60s"
authMethod:
iam:
awsAccessKeyId:
awsSecretAccessKey:
awsSecurityToken:
headerValue:
mount:
role:
gcp:
role:
mount:
credentials:
jwt_exp:
service_account:
project:
procPath: "/proc"
etcPath: "/etc"
varPath: "/var"
runPath: "/run"
sysPath: "/sys"