Update the fields in config.yaml to customize your Refinery configuration.
The default configuration at installation contains the minimum configuration needed to run Refinery.
Complete your Refinery set-up after configuring config.yaml by customizing your Sampling Methods configuration in rules.yaml.
config and rules.
We recommend using our Configuration Conversion Tool to migrate to Refinery 2.0.
The default Refinery configuration uses a hardcoded peer list for file-based peer management.
It uses the DeterministicSampler Sampling Method and a SampleRate of 1, meaning that no traffic will be dropped.
In the Refinery GitHub repository, a minimal default configuration file exists.
To see the full set of available options, refer below to the Refinery Config File. In the Refinery GitHub repository, a fully-commented configuration file version exists that can be used as a template.
These recommended settings are not required when using Refinery, but highlighted for their useful purposes.
To control access, use Network and Access Key configuration settings. For peer management, refer to Peer Management configuration, or if using Redis, Redis Peer Management.
If needed, Debugging contains configuration values used when setting up and debugging Refinery.
When configured, the Stress Relief mechanism can prevent Refinery from being overwhelmed by a large number of traces. It sheds load when Refinery is under duress and prevents crash loops.
Refinery supports using environment variables.
Environment variables take precedence over config.yaml file configuration.
| Environment Variable | Configuration Field |
|---|---|
REFINERY_GRPC_LISTEN_ADDRESS |
GRPCListenAddr |
REFINERY_REDIS_HOST |
PeerManagement.RedisHost |
REFINERY_REDIS_USERNAME |
PeerManagement.RedisUsername |
REFINERY_REDIS_PASSWORD |
PeerManagement.RedisPassword |
REFINERY_HONEYCOMB_API_KEY |
HoneycombLogger.LoggerAPIKey |
REFINERY_HONEYCOMB_METRICS_API_KEY REFINERY_HONEYCOMB_API_KEY |
LegacyMetrics.APIKey |
REFINERY_QUERY_AUTH_TOKEN |
QueryAuthToken |
REFINERY_HONEYCOMB_METRICS_API_KEY takes precedence over REFINERY_HONEYCOMB_API_KEY for the LegacyMetrics.APIKey configuration. The Refinery config file is a YAML file.
The file is split into sections; each section is a group of related configuration options.
Each section has a name, and the name is used to refer to the section in other parts of the config file.
This is an example config file:
General:
ConfigurationVersion: 2
Network:
ListenAddr: "0.0.0.0:8080"
PeerListenAddr: "0.0.0.0:8081"
OTelMetrics:
Enabled: true
APIKey: SetThisToAHoneycombKey
The remainder of this page describes the sections within the file and the fields in each.
General contains general configuration options that apply to the entire Refinery process.
ConfigurationVersion ConfigurationVersion is the file format of this particular configuration file.
This file is version 2. This field is required. It exists to allow the configuration system to adapt to future changes in the configuration file format.
int2MinRefineryVersion MinRefineryVersion is the minimum version of Refinery that can load this configuration file.
This setting specifies the lowest Refinery version capable of loading all of the features used in this file. If this value is present, then Refinery will refuse to start if its version is less than this setting.
stringv2.0DatasetPrefix DatasetPrefix is a prefix that can be used to distinguish a dataset from an environment in the rules.
If telemetry is being sent to both a classic dataset and a new environment called the same thing, such as production, then this parameter can be used to distinguish these cases.
When Refinery receives telemetry using an API key associated with a Honeycomb Classic dataset, it will then use the prefix in the form {prefix}. {dataset} when trying to resolve the rules definition.
stringConfigReloadInterval ConfigReloadInterval is the average interval between attempts at reloading the configuration file.
Refinery will attempt to read its configuration and check for changes at approximately this interval.
This time is varied by a random amount up to 10% to avoid all instances refreshing together.
In installations where configuration changes are handled by restarting Refinery, which is often the case when using Kubernetes, disable this feature with a value of 0s.
As of Refinery v2.7, news of a configuration change is immediately propagated to all peers, and they will attempt to reload their configurations.
Note that external factors (for example, Kubernetes ConfigMaps) may cause delays in propagating configuration changes.
duration15s Network contains network configuration options.
ListenAddr ListenAddr is the address where Refinery listens for incoming requests.
This setting is the IP and port on which Refinery listens for incoming HTTP requests.
These requests include traffic formatted as Honeycomb events, proxied requests to the Honeycomb API, and OpenTelemetry data using the http protocol.
Incoming traffic is expected to be HTTP, so if SSL is a requirement, put something like nginx in front to do the decryption.
hostport0.0.0.0:8080REFINERY_HTTP_LISTEN_ADDRESS--http-listen-addressPeerListenAddr PeerListenAddr is the IP and port on which to listen for traffic being rerouted from a peer.
Incoming traffic is expected to be HTTP, so if using SSL use something like nginx or a load balancer to do the decryption.
hostport0.0.0.0:8081REFINERY_PEER_LISTEN_ADDRESS--peer-listen-addressHTTPIdleTimeout HTTPIdleTimeout is the duration the http server waits for activity on the connection.
This is the amount of time after which if the http server does not see any activity, then it pings the client to see if the transport is still alive. “0s” means no timeout.
duration0sHoneycombAPI HoneycombAPI is the URL of the upstream Honeycomb API where the data will be sent.
This setting is the destination to which Refinery sends all events that it decides to keep.
urlhttps://api.honeycomb.ioREFINERY_HONEYCOMB_API--honeycomb-api AccessKeys contains access keys – API keys that the proxy will treat specially, and other flags that control how the proxy handles API keys.
ReceiveKeys ReceiveKeys is a set of Honeycomb API keys that the proxy will treat specially.
This list only applies to span traffic - other Honeycomb API actions will be proxied through to the upstream API directly without modifying keys.
stringarrayyour-key-goes-hereAcceptOnlyListedKeys AcceptOnlyListedKeys is a boolean flag that causes events arriving with API keys not in the ReceiveKeys list to be rejected.
If true, then only traffic using the keys listed in ReceiveKeys is accepted.
Events arriving with API keys not in the ReceiveKeys list will be rejected with an HTTP 401 error.
If false, then all traffic is accepted and ReceiveKeys is ignored.
This setting is applied before the SendKey and SendKeyMode settings.
boolSendKey SendKey is an optional Honeycomb API key that Refinery can use to send data to Honeycomb, depending on configuration. Setting this value via a command line flag may expose credentials - it is recommended to use the environment variable or a configuration file.
If SendKey is set to a valid Honeycomb key, then Refinery can use the listed key to send data.
The exact behavior depends on the value of SendKeyMode.
stringSetThisToAHoneycombKeyREFINERY_SEND_KEYSendKeyMode SendKeyMode controls how SendKey is used to replace or augment API keys used in incoming telemetry.
Controls how SendKey is used to replace or supply API keys used in incoming telemetry.
If AcceptOnlyListedKeys is true, then SendKeys will only be used for events with keys listed in ReceiveKeys.
none uses the incoming key for all telemetry (default).
all overwrites all keys, even missing ones, with SendKey.
nonblank overwrites all supplied keys but will not inject SendKey if the incoming key is blank.
listedonly overwrites only the keys listed in ReceiveKeys.
unlisted uses the SendKey for all events except those with keys listed in ReceiveKeys, which use their original keys.
missingonly uses the SendKey only to inject keys into events with blank keys.
All other events use their original keys.
stringnonenone, all, nonblank, listedonly, unlisted, missingonly RefineryTelemetry contains configuration information for the telemetry that Refinery uses to record its own operation.
AddRuleReasonToTrace AddRuleReasonToTrace controls whether to decorate traces with Refinery rule evaluation results.
When enabled, this setting causes traces that are sent to Honeycomb to include the field meta.refinery.reason.
This field contains text indicating which rule was evaluated that caused the trace to be included.
This setting also includes the field meta.refinery.send_reason, which contains the reason that the trace was sent.
Possible values of this field are trace_send_got_root, which means that the root span arrived; trace_send_expired, which means that TraceTimeout was reached; trace_send_ejected_full, which means that the trace cache was full; and trace_send_ejected_memsize, which means that Refinery was out of memory.
These names are also the names of metrics that Refinery tracks.
We recommend enabling this setting whenever a rules-based sampler is in use, as it is useful for debugging and understanding the behavior of your Refinery installation.
booltrueAddSpanCountToRoot AddSpanCountToRoot controls whether to add a metadata field to root spans that indicates the number of child elements in a trace.
The added metadata field, meta.span_count, indicates the number of child elements on the trace at the time the sampling decision was made.
This value is available to the rules-based sampler, making it possible to write rules that are dependent upon the number of spans, span events, and span links in the trace.
If true and AddCountsToRoot is set to false, then Refinery will add meta.span_count to the root span.
defaulttruetrueAddCountsToRoot AddCountsToRoot controls whether to add metadata fields to root spans that indicates the number of child spans, span events, span links, and honeycomb events.
If true, then Refinery will ignore the AddSpanCountToRoot setting and add the following fields to the root span based on the values at the time the sampling decision was made:
meta.span_count: the number of child spans on the trace
meta.span_event_count: the number of span events on the trace
meta.span_link_count: the number of span links on the trace
meta.event_count: the number of honeycomb events on the trace
Eligible for live reload.
Type: bool
AddHostMetadataToTrace AddHostMetadataToTrace specifies whether to add host metadata to traces.
If true, then Refinery will add the following tag to all traces: - meta.refinery.local_hostname: the hostname of the Refinery node
defaulttruetrue Traces contains configuration for how traces are managed.
SendDelay SendDelay is the duration to wait after the root span arrives before sending a trace.
This setting is a short timer that is triggered when a trace is marked complete by the arrival of the root span.
Refinery waits for this duration before sending the trace.
This setting exists to allow for asynchronous spans and small network delays to elapse before sending the trace.
SendDelay is not applied if the TraceTimeout expires or the SpanLimit is reached.
duration2sBatchTimeout BatchTimeout is how frequently Refinery sends unfulfilled batches.
By default, this setting uses the DefaultBatchTimeout in libhoney as its value, which is 100ms.
duration500msTraceTimeout TraceTimeout is the duration to wait before making the trace decision on an incomplete trace.
A long timer; it represents the outside boundary of how long to wait before making the trace decision about an incomplete trace. Normally trace decisions (send or drop) are made when the root span arrives. Sometimes the root span never arrives (for example, due to crashes). Once this timer fires, Refinery will make a trace decision based on the spans that have arrived so far. This ensures sending a trace even when the root span never arrives. After the trace decision has been made, Refinery retains a record of that decision for a period of time. When additional spans (including the root span) arrive, they will be kept or dropped based on the original decision. If particularly long-lived traces are present in your data, then you should increase this timer. Note that this increase will also increase the memory requirements for Refinery.
duration60sSpanLimit SpanLimit is the number of spans after which a trace becomes eligible for a trace decision.
This setting helps to keep memory usage under control. If a trace has more than this set number of spans, then it becomes eligible for a trace decision. It’s most helpful in a situation where a sudden burst of many spans in a large trace hits Refinery all at once, causing memory usage to spike and possibly crashing Refinery.
intMaxBatchSize MaxBatchSize is the maximum number of events to be included in each batch for sending.
This value is used to set the BatchSize field in the libhoney library used to send data to Honeycomb.
If you have particularly large traces, then you should increase this value.
Note that this will also increase the memory requirements for Refinery.
int500SendTicker SendTicker is the interval between checks for traces to send.
A short timer that determines the duration between trace cache review runs to send.
Increasing this will spend more time processing incoming events to reduce incoming_ or peer_router_dropped spikes.
Decreasing this will check the trace cache for timeouts more frequently.
duration100msMaxExpiredTraces MaxExpiredTraces is the maximum number of expired traces to process.
This setting controls how many traces are processed when it is time to make a sampling decision.
Up to this many traces will be processed every SendTicker duration.
If this number is too small it will mean Refinery is spending less time calculating sampling decisions, resulting in data arriving at Honeycomb slower.
If your collector_collect_loop_duration_ms is above 3 seconds it is recommended to reduce this value and the SendTicker duration.
This will mean Refinery makes fewer sampling decision calculations each SendTicker tick, but gets the chance to make decisions more often.
int5000 Debugging contains configuration values used when setting up and debugging Refinery.
DebugServiceAddr DebugServiceAddr is the IP and port where the debug service runs.
The debug service is generally only used when debugging Refinery itself, and will only run if the command line option -d is specified.
If this value is not specified, then the debug service runs on the first open port between localhost:6060 and localhost:6069.
hostportlocalhost:6060QueryAuthToken QueryAuthToken is the token that must be specified to access the /query endpoint. Setting this value via a command line flag may expose credentials - it is recommended to use the environment variable or a configuration file.
This token must be specified with the header “X-Honeycomb-Refinery-Query” in order for a /query request to succeed.
These /query requests are intended for debugging Refinery during setup and are not typically needed in normal operation.
If not specified, then the /query endpoints are inaccessible.
stringsome-private-valueREFINERY_QUERY_AUTH_TOKENAdditionalErrorFields AdditionalErrorFields is a list of span fields to include when logging errors happen during the ingestion of events.
For example, the span too large error.
This is primarily useful in trying to track down misbehaving senders in a large installation.
The fields dataset, apihost, and environment are always included.
If a field is not present in the span, then it will not be present in the error log.
stringarraytrace.span_idDryRun DryRun controls whether sampling is applied to incoming traces.
If enabled, then Refinery marks the traces that would be dropped given the current sampling rules, and sends all traces regardless of the sampling decision.
This is useful for evaluating sampling rules.
When DryRun is enabled, traces is decorated with meta.refinery. dryrun.kept that is set to true or false, based on whether the trace would be kept or dropped.
In addition, SampleRate will be set to the incoming rate for all traces, and the field meta.refinery.dryrun.sample_rate will be set to the sample rate that would have been used.
NOTE: This setting is not compatible with TraceCache=distributed, because drop trace decisions shared among peers do not contain all the relevant information needed to send traces to Honeycomb.
booltrue Logger contains configuration for logging.
Type Type is the type of logger to use.
The setting specifies where (and if) Refinery sends logs.
none means that logs are discarded.
honeycomb means that logs will be forwarded to Honeycomb as events according to the set Logging settings.
stdout means that logs will be written to stdout.
stringstdoutstdout, honeycomb, noneLevel Level is the logging level above which Refinery should send a log to the logger.
warn is the recommended level for production.
debug is very verbose, and should not be used in production environments.
stringwarndebug, info, warn, error, panic HoneycombLogger contains configuration for logging to Honeycomb.
Only used if Logger.Type is “honeycomb”.
APIHost APIHost is the URL of the Honeycomb API where Refinery sends its logs.
Refinery’s internal logs will be sent to this host using the standard Honeycomb Events API.
urlhttps://api.honeycomb.ioAPIKey APIKey is the API key used to send Refinery’s logs to Honeycomb. Setting this value via a command line flag may expose credentials - it is recommended to use the environment variable or a configuration file.
It is recommended that you create a separate team and key for Refinery logs.
stringSetThisToAHoneycombKeyREFINERY_HONEYCOMB_LOGGER_API_KEY, REFINERY_HONEYCOMB_API_KEYDataset Dataset is the dataset to which logs will be sent.
Only used if APIKey is specified.
stringRefinery LogsSamplerEnabled SamplerEnabled controls whether logs are sampled before sending to Honeycomb.
The sample rate is controlled by the SamplerThroughput setting.
The sampler used throttles the rate of logs sent to Honeycomb from any given source within Refinery – it should effectively limit the rate of redundant messages.
defaulttruetrueSamplerThroughput SamplerThroughput is the sampling throughput for logs in events per second.
The sampling algorithm attempts to make sure that the average throughput approximates this value, while also ensuring that all unique logs arrive at Honeycomb at least once per sampling period.
float1010 StdoutLogger contains configuration for logging to stdout.
Only used if Logger.Type is “stdout”.
Structured Structured controls whether to use structured logging.
true generates structured logs (JSON).
false generates plain text logs.
boolSamplerEnabled SamplerEnabled controls whether logs are sampled before sending to stdout.
The sample rate is controlled by the SamplerThroughput setting.
boolSamplerThroughput SamplerThroughput is the sampling throughput for logs in events per second.
The sampling algorithm attempts to make sure that the average throughput approximates this value, while also ensuring that all unique logs arrive at stdout at least once per sampling period.
float1010 PrometheusMetrics contains configuration for Refinery’s internally-generated metrics as made available through Prometheus.
Enabled Enabled controls whether to expose Refinery metrics over the PrometheusListenAddr port.
Each of the metrics providers can be enabled or disabled independently. Metrics can be sent to multiple destinations.
boolListenAddr ListenAddr is the IP and port the Prometheus Metrics server will run on.
Determines the interface and port on which Prometheus will listen for requests for /metrics.
Must be different from the main Refinery listener.
Only used if Enabled is true in PrometheusMetrics.
hostportlocalhost:2112 LegacyMetrics contains configuration for Refinery’s legacy metrics.
Version 1.x of Refinery used this format for sending Metrics to Honeycomb.
The metrics generated that way are nonstandard and will be deprecated in a future release.
New installations should prefer OTelMetrics.
Enabled Enabled controls whether to send legacy-formatted metrics to Honeycomb.
Each of the metrics providers can be enabled or disabled independently. Metrics can be sent to multiple destinations.
boolAPIHost APIHost is the URL of the Honeycomb API where legacy metrics are sent.
Refinery’s internal metrics will be sent to this host using the standard Honeycomb Events API.
urlhttps://api.honeycomb.ioAPIKey APIKey is the API key used by Refinery to send its metrics to Honeycomb. Setting this value via a command line flag may expose credentials - it is recommended to use the environment variable or a configuration file.
It is recommended that you create a separate team and key for Refinery metrics.
stringSetThisToAHoneycombKeyREFINERY_HONEYCOMB_METRICS_API_KEY, REFINERY_HONEYCOMB_API_KEYDataset Dataset is the Honeycomb dataset where Refinery sends its metrics.
Only used if APIKey is specified.
stringRefinery MetricsReportingInterval ReportingInterval is the interval between sending legacy metrics to Honeycomb.
Between 1 and 60 seconds is typical.
duration30s OTelMetrics contains configuration for Refinery’s OpenTelemetry (OTel) metrics.
This is the preferred way to send metrics to Honeycomb.
New installations should prefer OTelMetrics.
Enabled Enabled controls whether to send metrics via OpenTelemetry.
Each of the metrics providers can be enabled or disabled independently. Metrics can be sent to multiple destinations.
boolAPIHost APIHost is the URL of the OpenTelemetry API to which metrics will be sent.
Refinery’s internal metrics will be sent to the /v1/metrics endpoint on this host.
urlhttps://api.honeycomb.ioAPIKey APIKey is the API key used to send Honeycomb metrics via OpenTelemetry. Setting this value via a command line flag may expose credentials - it is recommended to use the environment variable or a configuration file.
It is recommended that you create a separate team and key for Refinery metrics.
If this is blank, then Refinery will not set the Honeycomb-specific headers for OpenTelemetry, and your APIHost must be set to a valid OpenTelemetry endpoint.
stringSetThisToAHoneycombKeyREFINERY_OTEL_METRICS_API_KEY, REFINERY_HONEYCOMB_API_KEYDataset Dataset is the Honeycomb dataset that Refinery sends its OpenTelemetry metrics.
Only used if APIKey is specified.
stringRefinery MetricsReportingInterval ReportingInterval is the interval between sending OpenTelemetry metrics to Honeycomb.
Between 1 and 60 seconds is typical.
duration30sCompression Compression is the compression algorithm to use when sending OpenTelemetry metrics to Honeycomb.
gzip is the default and recommended value.
In rare circumstances, compression costs may outweigh the benefits, in which case none may be used.
stringgzipnone, gzip OTelTracing contains configuration for Refinery’s own tracing.
Enabled Enabled controls whether to send Refinery’s own OpenTelemetry traces.
The setting specifies if Refinery sends traces.
boolAPIHost APIHost is the URL of the OpenTelemetry API to which traces will be sent.
Refinery’s internal traces will be sent to the /v1/traces endpoint on this host.
urlhttps://api.honeycomb.ioAPIKey APIKey is the API key used to send Refinery’s traces to Honeycomb. Setting this value via a command line flag may expose credentials - it is recommended to use the environment variable or a configuration file.
It is recommended that you create a separate team and key for Refinery telemetry.
If this value is blank, then Refinery will not set the Honeycomb-specific headers for OpenTelemetry, and your APIHost must be set to a valid OpenTelemetry endpoint.
stringSetThisToAHoneycombKeyREFINERY_HONEYCOMB_TRACES_API_KEY, REFINERY_HONEYCOMB_API_KEYDataset Dataset is the Honeycomb dataset to which Refinery sends its OpenTelemetry metrics.
Only used if APIKey is specified.
stringRefinery TracesSampleRate SampleRate is the rate at which Refinery samples its own traces.
This is the Honeycomb sample rate used to sample traces sent by Refinery.
Since each incoming span generates multiple outgoing spans, a minimum sample rate of 100 is strongly advised.
int100Insecure Insecure controls whether to send Refinery’s own OpenTelemetry traces via http instead of https.
When true Refinery will export its internal traces over http instead of https. Useful if you plan on sending your traces to a different refinery instance for tail sampling.
bool PeerManagement controls how the Refinery cluster communicates between peers.
Type Type is the type of peer management to use.
Peer management is the mechanism by which Refinery locates its peers.
file means that Refinery gets its peer list from the Peers list in this config file.
It also prevents Refinery from using a publish/subscribe mechanism to propagate peer lists, stress levels, and configuration changes.
redis means that Refinery uses a Publish/Subscribe mechanism, implemented on Redis, to propagate peer lists, stress levels, and notification of configuration changes much more quickly than the legacy mechanism.
The recommended setting is redis, especially for new installations.
If redis is specified, fields in RedisPeerManagement must also be set.
stringfileredis, fileIdentifier Identifier specifies the identifier to use when registering itself with peers.
By default, when using a peer registry, Refinery will use the local hostname to identify itself to other peers.
If your environment requires something else, (for example, if peers cannot resolve each other by name), then you can specify the exact identifier, such as an IP address, to use here.
Overrides IdentifierInterfaceName, if both are set.
string192.168.1.1IdentifierInterfaceName IdentifierInterfaceName specifies a network interface to use when finding a local hostname.
By default, when using a peer registry, Refinery will use the local hostname to identify itself to other peers. If your environment requires that you use IPs as identifiers (for example, if peers cannot resolve each other by name), then you can specify the network interface that Refinery is listening on here. Refinery will use the first unicast address that it finds on the specified network interface as its identifier.
stringeth0UseIPV6Identifier UseIPV6Identifier specifies that Refinery should use an IPV6 address as its identifier.
If using IdentifierInterfaceName, Refinery will default to the first IPv4 unicast address it finds for the specified interface.
If this value is specified, then Refinery will use the first IPV6 unicast address found.
boolPeers Peers is the list of peers to use when Type is “file”, excluding self.
This list is ignored when Type is “redis”. The format is a list of strings of the form “scheme://host:port”.
stringarrayhttp://192.168.1.11:8081,http://192.168.1.12:8081 RedisPeerManagement controls how the Refinery cluster communicates between peers when using Redis.
Does not apply when PeerManagement.Type is “file”.
Host Host is the host and port of the Redis instance to use for peer cluster membership management.
Must be in the form host:port.
hostportlocalhost:6379REFINERY_REDIS_HOSTClusterHosts ClusterHosts is a list of host and port pairs for the instances in a Redis Cluster, and used for managing peer cluster membership.
This configuration enables Refinery to connect to a Redis deployment setup in Cluster Mode.
Each entry in the list should follow the format host:port.
If ClusterHosts is specified, the Host setting will be ignored.
stringarray- localhost:6379Username Username is the username used to connect to Redis for peer cluster membership management. Setting this value via a command line flag may expose credentials - it is recommended to use the environment variable or a configuration file.
Many Redis installations do not use this field.
stringREFINERY_REDIS_USERNAMEPassword Password is the password used to connect to Redis for peer cluster membership management. Setting this value via a command line flag may expose credentials - it is recommended to use the environment variable or a configuration file.
Many Redis installations do not use this field.
stringREFINERY_REDIS_PASSWORDAuthCode AuthCode is the string used to connect to Redis for peer cluster membership management using an explicit AUTH command. Setting this value via a command line flag may expose credentials - it is recommended to use the environment variable or a configuration file.
Many Redis installations do not use this field.
stringREFINERY_REDIS_AUTH_CODEUseTLS UseTLS enables TLS when connecting to Redis for peer cluster membership management.
When enabled, this setting sets the MinVersion in the TLS configuration to 1.2.
boolUseTLSInsecure UseTLSInsecure disables certificate checks when connecting to Redis for peer cluster membership management.
This setting is intended for use with self-signed certificates and sets the InsecureSkipVerify flag within Redis.
boolTimeout Timeout is the timeout to use when communicating with Redis.
It is rarely necessary to adjust this value.
duration5s Collection contains the settings that are relevant to collecting spans together to make traces.
If none of the memory settings are used, then Refinery will not attempt to limit its memory usage.
This is not recommended for production use since a burst of traffic could cause Refinery to run out of memory and crash.
CacheCapacity CacheCapacity is the number of traces to keep in the cache’s circular buffer.
The collection cache is used to collect all active spans into traces.
It is organized as a circular buffer.
When the buffer wraps around, Refinery will try a few times to find an empty slot; if it fails, it starts ejecting traces from the cache earlier than would otherwise be necessary.
Ideally, the size of the cache should be many multiples (100x to 1000x) of the total number of concurrently active traces (average trace throughput * average trace duration).
NOTE: This setting is now deprecated and no longer controls the cache size.
Instead the maxmimum memory usage is controlled by MaxMemoryPercentage and MaxAlloc.
int10000PeerQueueSize PeerQueueSize is the maximum number of in-flight spans redirected from other peers stored in the peer span queue.
The peer span queue serves as a buffer for spans redirected from other peers before they are processed.
In the event that this queue reaches its capacity, any subsequent spans will be discarded.
The size of this queue is contingent upon the number of peers within the cluster.
Specifically, with N peers, the queue’s span capacity is determined by (N-1)/N of the total number of spans.
Its minimum value should be at least three times the CacheCapacity.
int30000IncomingQueueSize IncomingQueueSize is the number of in-flight spans to keep in the incoming span queue.
The incoming span queue is used to buffer spans before they are processed.
If this queue fills up, then subsequent spans will be dropped.
Its minimum value should be at least three times the CacheCapacity.
int30000AvailableMemory AvailableMemory is the amount of system memory available to the Refinery process.
This value will typically be set through an environment variable controlled by the container or deploy script.
If this value is zero or not set, then MaxMemoryPercentage cannot be used to calculate the maximum allocation and MaxAlloc will be used instead.
If set, then this must be a memory size.
Sizes with standard unit suffixes (such as MB and GiB) and Kubernetes units (such as M and Gi) are supported.
Fractional values with a suffix are supported.
If AvailableMemory is set, Collections.MaxAlloc must not be defined.
memorysize4.5GbREFINERY_AVAILABLE_MEMORY--available-memoryMaxMemoryPercentage MaxMemoryPercentage is the maximum percentage of memory that should be allocated by the span collector.
If nonzero, then it must be an integer value between 1 and 100, representing the target maximum percentage of memory that should be allocated by the span collector.
If set to a non-zero value, then once per tick (see SendTicker) the collector will compare total allocated bytes to this calculated value.
If allocation is too high, then traces will be ejected from the cache early to reduce memory.
Useful values for this setting are generally in the range of 70-90.
percentage7575MaxAlloc MaxAlloc is the maximum number of bytes that should be allocated by the Collector.
If set, then this must be a memory size.
Sizes with standard unit suffixes (such as MB and GiB) and Kubernetes units (such as M and Gi) are supported.
Fractional values with a suffix are supported.
See MaxMemoryPercentage for more details.
If set, Collections.AvailableMemory must not be defined.
memorysizeDisableRedistribution DisableRedistribution controls whether to transmit traces in cache to remaining peers during cluster scaling event.
If true, Refinery will NOT forward live traces in its cache to the rest of the peers when peers join or leave the cluster.
By disabling this behavior, it can help to prevent disruptive bursts of network traffic when large traces with long TraceTimeout are redistributed.
boolRedistributionDelay RedistributionDelay controls the amount of time Refinery waits after each cluster scaling event before redistributing in-memory traces.
This value should be longer than the amount of time between individual pod changes in a bulk scaling operation (changing the cluster size by more than one pod). Each redistribution generates additional traffic between peers. If this value is too short, multiple consecutive redistributions will occur and the resulting traffic may overwhelm the cluster.
duration30sShutdownDelay ShutdownDelay controls the maximum time Refinery can use while draining traces at shutdown.
This setting controls the duration that Refinery expects to have to drain in-process traces before shutting down an instance. When asked to shut down gracefully, Refinery stops accepting new spans immediately and drains the remaining traces by sending them to remaining peers. This value should be set to a bit less than the normal timeout period for shutting down without forcibly terminating the process.
duration15sTraceLocalityMode TraceLocalityMode controls how Refinery handles spans that belong to the same trace in a clustered environment.
When concentrated, Refinery will route all spans that belong to the same trace to a single peer.
This is the default behavior (“Trace Locality”) and the way Refinery has worked in the past.
When distributed, Refinery will instead keep spans on the node where they were received, and forward proxy spans that contain only the key information needed to make a trace decision.
This can reduce the amount of traffic between peers, and can help avoid a situation where a single large trace can cause a memory overrun on a single node.
If distributed, the amount of traffic between peers will be reduced, but the amount of traffic between Refinery and Redis will significantly increase, because Refinery uses Redis to distribute the trace decisions to all nodes in the cluster.
It is important to adjust the size of the Redis cluster in this case.
The total volume of network traffic in distributed mode should be expected to decrease unless the cluster size is very large (hundreds of nodes).
NOTE: This setting is not compatible with DryRun when set to distributed.
See DryRun for more information.
stringconcentratedHealthCheckTimeout HealthCheckTimeout controls the maximum duration allowed for collection health checks to complete.
The HealthCheckTimeout setting specifies the maximum duration allowed for the health checks of the collection subsystems to complete.
If a subsystem does not respond within this timeout period, it will be marked as unhealthy.
This timeout value should be set carefully to ensure that transient delays do not lead to unnecessary failure detection while still allowing for timely identification of actual health issues.
duration3s BufferSizes contains the settings that are relevant to the sizes of communications buffers.
UpstreamBufferSize UpstreamBufferSize is the size of the queue used to buffer spans to send to the upstream Collector.
The size of the buffer is measured in spans. If the buffer fills up, then performance will degrade because Refinery will block while waiting for space to become available. If this happens, then you should increase the buffer size.
int10000PeerBufferSize PeerBufferSize is the size of the queue used to buffer spans to send to peer nodes.
The size of the buffer is measured in spans. If the buffer fills up, then performance will degrade because Refinery will block while waiting for space to become available. If this happens, then you should increase this buffer size.
int100000 Specialized contains special-purpose configuration options that are not typically needed.
EnvironmentCacheTTL EnvironmentCacheTTL is the duration for which environment information is cached.
This is the amount of time for which Refinery caches environment information, which it looks up from Honeycomb for each different APIKey.
This information is used when making sampling decisions.
If you have a very large number of environments, then you may want to increase this value.
duration1hCompressPeerCommunication CompressPeerCommunication determines whether Refinery will compress span data it forwards to peers.
If it costs money to transmit data between Refinery instances (for example, when spread across AWS availability zones), then you almost certainly want compression enabled to reduce your bill. The option to disable it is provided as an escape hatch for deployments that value lower CPU utilization over data transfer costs.
defaulttruetrueAdditionalAttributes AdditionalAttributes is a map that can be used for injecting user-defined attributes into every span.
For example, it could be used for naming a Refinery cluster. Both keys and values must be strings.
mapClusterName:MyCluster,environment:production IDFields controls the field names to use for the event ID fields.
These fields are used to identify events that are part of the same trace.
TraceNames TraceNames is the list of field names to use for the trace ID.
The first field in the list that is present in an incoming span will be used as the trace ID. If none of the fields are present, then Refinery treats the span as not being part of a trace and forwards it immediately to Honeycomb.
stringarraytrace.trace_id,traceIdParentNames ParentNames is the list of field names to use for the parent ID.
The first field in the list that is present in an event will be used as the parent ID.
A trace without a parent_id is assumed to be a root span.
stringarraytrace.parent_id,parentId GRPCServerParameters controls the parameters of the gRPC server used to receive OpenTelemetry data in gRPC format.
Enabled Enabled specifies whether the gRPC server is enabled.
If false, then the gRPC server is not started and no gRPC traffic is accepted.
defaulttruetrueListenAddr ListenAddr is the address Refinery listens to for incoming GRPC OpenTelemetry events.
Incoming traffic is expected to be unencrypted, so if using SSL, then put something like nginx in front to do the decryption.
hostportREFINERY_GRPC_LISTEN_ADDRESS--grpc-listen-addressMaxConnectionIdle MaxConnectionIdle is the amount of time to permit an idle connection.
A duration for the amount of time after which an idle connection will be closed by sending a GoAway. “Idle” means that there are no active RPCs. “0s” sets duration to infinity, but this is not recommended for Refinery deployments behind a load balancer, because it will prevent the load balancer from distributing load evenly among peers.
duration1m1mMaxConnectionAge MaxConnectionAge is the maximum amount of time a gRPC connection may exist.
After this duration, the gRPC connection is closed by sending a GoAway.
A random jitter of +/-10% will be added to MaxConnectionAge to spread out connection storms.
0s sets duration to infinity; a value measured in low minutes will help load balancers to distribute load among peers more evenly.
duration3mMaxConnectionAgeGrace MaxConnectionAgeGrace is the duration beyond MaxConnectionAge after which the connection will be forcibly closed.
This setting is in case the upstream node ignores the GoAway request.
“0s” sets duration to infinity.
duration1mKeepAlive KeepAlive is the duration between keep-alive pings.
After this amount of time, if the client does not see any activity, then it pings the server to see if the transport is still alive. “0s” sets duration to 2 hours.
duration1mKeepAliveTimeout KeepAliveTimeout is the duration the server waits for activity on the connection.
This is the amount of time after which if the server does not see any activity, then it pings the client to see if the transport is still alive. “0s” sets duration to 20 seconds.
duration20sMaxSendMsgSize MaxSendMsgSize is the maximum message size the server can send.
The server enforces a maximum message size to avoid exhausting the memory available to the process by a single request. The size is expressed in bytes.
memorysize15MBMaxRecvMsgSize MaxRecvMsgSize is the maximum message size the server can receive.
The server enforces a maximum message size to avoid exhausting the memory available to the process by a single request. The size is expressed in bytes.
memorysize15MB SampleCache controls the sample cache used to retain information about trace status after the sampling decision has been made.
KeptSize KeptSize is the number of traces preserved in the cuckoo kept traces cache.
Refinery keeps a record of each trace that was kept and sent to Honeycomb, along with some statistical information.
This is most useful in cases where the trace was sent before sending the root span, so that the root span can be decorated with accurate metadata.
Default is 10_000 traces.
Each trace in this cache consumes roughly 200 bytes.
int10000DroppedSize DroppedSize is the size of the cuckoo dropped traces cache.
This cache consumes 4-6 bytes per trace at a scale of millions of traces. Changing its size with live reload sets a future limit, but does not have an immediate effect.
int1000000SizeCheckInterval SizeCheckInterval controls how often the cuckoo cache re-evaluates its remaining capacity.
This cache is quite resilient so it does not need to happen very often, but the operation is also inexpensive. Default is 10 seconds.
duration10s StressRelief controls the Stress Relief mechanism, which is used to prevent Refinery from being overwhelmed by a large number of traces.
There is a metric called stress_level that is emitted as part of Refinery metrics.
It is a measure of Refinery’s throughput rate relative to its processing rate, combined with the amount of room in its internal queues, and ranges from 0 to 100.
stress_level is generally expected to be 0 except under heavy load.
When stress levels reach 100, there is an increased chance that Refinery will become unstable.
To avoid this problem, the Stress Relief system can do deterministic sampling on new trace traffic based solely on TraceID, without having to store traces in the cache or take the time processing sampling rules.
Existing traces in flight will be processed normally, but when Stress Relief is active, trace decisions are made deterministically on a per-span basis; all spans will be sampled according to the SamplingRate specified here.
Once Stress Relief activates (by exceeding the ActivationLevel), it will not deactivate until stress_level falls below the DeactivationLevel.
When it deactivates, normal trace decisions are made – and any additional spans that arrive for traces that were active during Stress Relief will respect the decisions made during that time.
The measurement of stress is a lagging indicator and is highly dependent on Refinery configuration and scaling.
Other configuration values should be well tuned first, before adjusting the Stress Relief Activation parameters.
Stress Relief is not a substitute for proper configuration and scaling, but it can be used as a safety valve to prevent Refinery from becoming unstable under heavy load.
Mode Mode is a string indicating how to use Stress Relief.
This setting sets the Stress Relief mode.
“never” means that Stress Relief will never activate.
“monitor” is the recommended setting, and means that Stress Relief will monitor the status of Refinery and activate according to the levels set by fields such as ActivationLevel.
“always” means that Stress Relief is always on, which may be useful in an emergency situation.
stringneverActivationLevel ActivationLevel is the stress_level (from 0-100) at which Stress Relief is triggered.
This value must be greater than DeactivationLevel and should be high enough that it is not reached in normal operation.
percentage90DeactivationLevel DeactivationLevel is the stress_level (from 0-100) at which Stress Relief is turned off.
This setting is subject to MinimumActivationDuration.
The value must be less than ActivationLevel.
percentage75SamplingRate SamplingRate is the sampling rate to use when Stress Relief is activated.
All new traces will be deterministically sampled at this rate based only on the traceID.
It should be chosen to be a rate that sends fewer samples than the average sampling rate Refinery is expected to generate.
For example, if Refinery is configured to normally sample at a rate of 1 in 10, then Stress Relief should be configured to sample at a rate of at least 1 in 30.
int100MinimumActivationDuration MinimumActivationDuration is the minimum time that Stress Relief will stay enabled once activated.
This setting helps to prevent oscillations.
duration10s