Probes Config

# Value can either be a static value or can be derived from target's labels. # To get value from target's labels, use target.labels.<target's label key> # as value.

# Whether to log metrics or not.

# Probe name. It should be unique across all probes.

# Interval between two probe runs in milliseconds. # Only one of "interval" and "inteval_msec" should be defined. # Default interval is 2s.

# Interval between two probe runs in string format, e.g. 10s. # Only one of "interval" and "inteval_msec" should be defined. # Default interval is 2s.

# Timeout for each probe in milliseconds # Only one of "timeout" and "timeout_msec" should be defined. # Default timeout is 1s.

# Timeout for each probe in string format, e.g. 10s. # Only one of "timeout" and "timeout_msec" should be defined. # Default timeout is 1s.

# Targets for the probe. Targets are required for all probes except # for external, user_defined, and extension probe types.

# Latency distribution. If specified, latency is stored as a distribution.

# Latency unit. Any string that's parseable by time.ParseDuration. # Valid values: "ns", "us" (or "µs"), "ms", "s", "m", "h".

# Latency metric name. You may want to change the latency metric name, if: # you're using latency_distribution for some probes, and regular metric for # other probes, and you want to differentiate between the two. # For example: # probe { # name: "web1_latency" # latency_distribution: {...} # latency_metric_name: "latency_dist" # ... # } # probe { # name: "app1" # ... # }

# Validators for this probe. Validators are run on the data returned by the # probe. See https://cloudprober.org/docs/how-to/validators/ for more info.

# Set the source IP to send packets from, either by providing an IP address # directly, or a network interface.

# How often to export stats. Probes usually run at a higher frequency (e.g. # every second); stats from individual probes are aggregated within # cloudprober until exported. In most cases, users don't need to change the # default. # # By default this field is set in the following way: # For all probes except UDP: # stats_export_interval=max(interval, 10s) # For UDP: # stats_export_interval=max(2*max(interval, timeout), 10s)

# Additional labels to add to the probe results. Label's value can either be # static or can be derived from target's labels. # # Example: # additional_label { # key: "app" # value: "@target.label.app@" # } # (More detailed example at: examples/additional_label/cloudprober.cfg)

# (Experimental) If set, test is inversed, i.e. we count it as success if # target doesn't respond. This is useful, for example, that your firewall is # working as expected. # # This is currently implemented only by PING and TCP probes. # Note: This field is currently experimental, and may change in future.

# Alerts configuration. If specified, cloudprober will generate alerts on # probe failures. You can specify multiple alerts. # Example: # alert { # name: "alert1" # condition {...} # notify { # pagerduty { ...} # } # } # alert { # name: "alert2" # notify { ... } # }

# Which machines this probe should run on. If defined, cloudprober will run # this probe only if machine's hostname matches this value. This is useful # for large deployments, where you may want to use the same prober config # everywhere but run this probe only on a subset of machines.

# Schedule for the probe. You can use a schedule to specify when a probe # should or should not run. This is useful for running probes only during # business hours. # # You can specify multiple schedules. Probe will not run if any of the # "DISABLE" schedules are active. If both "ENABLE" and "DISABLE" schedules # overlap, "DISABLE" takes precedence. # # For example, to disable a probe during weekends and on Tuesday between 7pm # and 8pm, e.g. for rollouts: # schdule { # type: DISABLE # start_weekday: FRIDAY # start_time: "20:00" # end_weekday: SUNDAY # end_time: "17:00" # timezone: "America/New_York" # } # schdule { # type: DISABLE # start_weekday: TUESDAY # start_time: "19:00" # end_weekday: TUESDAY # end_time: "20:00" # timezone: "America/New_York" # }

# Debug options. Currently only used to enable logging metrics.

# Period start weekday. If not specified, it defaults to EVERYDAY.

# Start time in 24 hour HH:MM format.

# Period end weekday. If not specified, it defaults to EVERYDAY.

# End time in 24 hour HH:MM format.

# Timezone in which the probe should run. If not specified, it defaults to # UTC. Example: "America/New_York"

# Azure account name and key. If you want to use managed identities, leave # account_key empty.

# Azure endpoint. Default is "https://<account>.blob.core.windows.net".

# OAuth2 configuration. If you want to use managed identities, leave this # field empty. See # https://cloudprober.org/docs/config/latest/oauth/#cloudprober_oauth_Config # for more details on oauth.Config.

# Serve test artifacts on Cloudprober's default webserver. This is # disabled by default for security reasons.

# Specify web server path to serve test artifacts on. # Default is "/artifacts/<probename>".

# Storage for test artifacts. Note that test artifacts are always # written to the workdir first, and uploaded to the storage backend in a # parallel goroutine. This is to make sure that uploads don't block the # main probe execution.

# Maximum age of artifacts in seconds.

# Cleanup interval in seconds. Default is 1 hour or max_age_sec, whichever # is smaller.

# If you want to use default credentials on GCE or GKE, leave this field # empty. See # https://cloudprober.org/docs/config/latest/oauth/#cloudprober_oauth_GoogleCredentials # for more details on oauth.GoogleCredentials.

# GCS endpoint.

# Cleanup options for local storage. Specifying cleanup options for local # enables automatic cleanup of old artifacts.

# Playwright test specs. These are passed to playwright as it is. This # field works in conjunction with test_dir -- test specs should be under # test directory. # # If test_spec is not specified, all test specs in the test directory are # executed, and since default test directory is config file's directory, # if you leave both the fields unspecified, all test specs co-located with # the config file are executed.

# Test directory. This is the directory where test specs are located. # Default test_dir is config file directory ("{{configDir}}").

# Workdir is path to the working directory. It should be writable. If not # specified, we try to create a temporary directory. All the output files # and reports are stored under <workdir>/output/. # If you need to be able access the output files, you should set this # field to a persistent location, e.g. a persistent volume, or configure # artifact options.

# Path to the playwright installation. We execute tests from this location. # If not specified, we'll use the value of environment variable # $PLAYWRIGHT_DIR, which is automatically set by the official cloudprober # playwright image (tag: "<version>-pw").

# NPX path. Default is to assume npx is in the PATH.

# Whether to enable screenshots for successful tests as well. # Note that screenshots are always enabled for failed tests, and you can # always save screenshots explicitly in the test spec.

# Traces are expensive and can slow down the test. We recommend to enable # this only when needed.

# By default, we export all test metrica as counters. You can change how # metrics are exported by setting the following options.

# Artifacts options.

# Cleanup options.

# Environment variables. These are passed/set before probing starts.

# Requests per probe. # Number of DNS requests per probe. Requests are executed concurrently and # each DNS request contributes to probe results. For example, if you run two # requests per probe, "total" counter will be incremented by 2.

# How long to wait between two requests to the same target. Only relevant # if requests_per_probe is also configured. # # This value should be less than (interval - timeout) / requests_per_probe. # This is to ensure that all requests are executed within one probe interval # and all of them get sufficient time. For example, if probe interval is 2s, # timeout is 1s, and requests_per_probe is 10, requests_interval_msec # should be less than 10ms.

# S3 endpoint. If not specified, default endpoint for the region is used.

# Domain to use when making DNS queries

# DNS Query Type

# Minimum number of answers expected. Default behavior is to return success # if DNS response status is NOERROR.

# Whether to resolve the target (target is DNS server here) before making # the request. If set to false, we hand over the target directly to the DNS # client. Otherwise, we resolve the target first to an IP address. By # default we resolve first if it's a discovered resource, e.g., a k8s # endpoint.

# DNS Query QueryClass

# Which DNS protocol is used for resolution.

# Requests per probe. # Number of DNS requests per probe. Requests are executed concurrently and # each DNS request contributes to probe results. For example, if you run two # requests per probe, "total" counter will be incremented by 2.

# How long to wait between two requests to the same target. Only relevant # if requests_per_probe is also configured. # # This value should be less than (interval - timeout) / requests_per_probe. # This is to ensure that all requests are executed within one probe interval # and all of them get sufficient time. For example, if probe interval is 2s, # timeout is 1s, and requests_per_probe is 10, requests_interval_msec # should be less than 10ms.

# Command. For ONCE probes, arguments are processed for the following field # substitutions: # @probe@ Name of the probe # @target.name@ or @target@ Hostname of the target # @target.port@ or @port@ Port of the target # @target.ip@ IP address associated with target # @address@ Resolved IP address of the target, in case of # discovered targets, same as @target.ip@. # @target.label.<x>@ Label x of the target # # For example, for target ig-us-central1-a, /tools/recreate_vm -vm @target@ # will get converted to: /tools/recreate_vm -vm ig-us-central1-a

# Command environment variables. These are passed on to the external probe # process as environment variables.

# Export output as metrics, where output is the output returned by the # external probe process, over stdout for ONCE probes, and through ProbeReply # for SERVER probes. Cloudprober expects variables to be in the following # format in the output: # var1 value1 (for example: total_errors 589)

# (Only applicable to ONCE mode). Disable streaming output metrics. By # default, external probe will export output metrics as they are available # on the stdout. If this option is set to true, output metrics will be # exported only after the probe has completed. # New in version 0.13.4. This was true by default in previous versions.

# Protoset contains descriptor source protos generated from the *.proto # files. You can use protoc to generate protoset files: # protoc --proto_path=. --descriptor_set_out=myservice.protoset \ # --include_imports my/custom/server/service.proto

# Note first 3 methods are valid only if descriptor source is not set.

# Request data (in JSON format) for the call_service_method request.

# Port for gRPC requests (Corresponding target field: port) # Default is 443, but if this field is not set and target has a port, either # discovered (e.g., k8s services, ingresses), or configured (e.g. endpoint, # file targets), we use target's port.

# Optional oauth config. For GOOGLE_DEFAULT_CREDENTIALS, use: # oauth_config: { bearer_token { gce_service_account: "default" } }

# If alts_config is provided, gRPC client uses ALTS for authentication and # encryption. For default alts configs, use: # alts_config: {}

# If TLSConfig is specified, it's used for authentication. # Note that only one of ALTSConfig and TLSConfig can be enabled at a time.

# if insecure_transport is set to true, TLS will not be used.

# Blob size for ECHO, READ, and WRITE methods.

# For HEALTH_CHECK, name of the service to health check.

# For HEALTH_CHECK, ignore status. By default, HEALTH_CHECK test passes # only if response-status is SERVING. Setting the following option makes # HEALTH_CHECK pass regardless of the response-status.

# Request definition for the GENERIC method.

# Number of connections to use. Default is 2 for ECHO, READ and WRITE # methods for backward compatibility. For HEALTH_CHECK and GENERIC, default # is 1.

# If connect_timeout is not specified, reuse probe timeout. Note that this # timeout will have an impact only if it is less than the probe timeout.

# URI scheme allows gRPC to use different resolvers # Example URI scheme: "google-c2p:///" # See https://github.com/grpc/grpc/blob/master/doc/naming.md for more details

# If provided, ALTS verifies that peer is using one of the given service # accounts.

# Handshaker service address. Default is to use the local metadata server. # For most of the ALTS use cases, default address should be okay.

# HTTP request scheme (Corresponding target label: "scheme"). If not set, we # use taget's 'scheme' label if present. # Note: protocol is deprecated, use scheme instead.

# Relative URL (Corresponding target label: "path"). We construct the final # URL like this: # <scheme>://<host>:<port>/<relative_url>. # # Note that the relative_url should start with a '/'.

# Port for HTTP requests (Corresponding target field: port) # Default is to use the scheme specific port, but if this field is not # set and discovered target has a port (e.g., k8s services, ingresses), # we use target's port.

# Whether to resolve the target before making the request. If set to true, # we resolve the target first to an IP address and make a request using # that while passing target name (or 'host' label if present) as Host # header. # # This behavior is automatic for discovered targets if they have an IP # address associated with them. Usually you don't need to worry about this # field and you can left it unspecified. We'll ty to do the right thing.

# Export response (body) count as a metric

# HTTP request method

# HTTP request headers # It is recommended to use "header" instead of "headers" for new configs. # header { # key: "Authorization" # value: "Bearer {{env "AUTH_TOKEN"}}" # }

# Request body. This field works similar to the curl's data flag. If there # are multiple "body" fields, we combine their values with a '&' in between. # # Also, we try to guess the content-type header based on the data: # 1) If data appears to be a valid json, we automatically set the # content-type header to "application/json". # 2) If the final data string appears to be a valid query string, we # set content-type to "application/x-www-form-urlencoded". Content type # header can still be overridden using the header field above. # Example: # body: "grant_type=client_credentials" # body: "scope=transferMoney" # body: "clientId=aweseomeClient" # body: "clientSecret=noSecret"

# Request body from file. This field is similar to the body field above, but # value is read from a file. # TODO(manugarg): We should consider providing a way to substitute environment # variables in the file.

# Enable HTTP keep-alive. If set to true, underlying connection is reused # for further probes. Default is to close the connection after every request.

# OAuth Config

# Disable HTTP2 # Golang HTTP client automatically enables HTTP/2 if server supports it. This # option disables that behavior to enforce HTTP/1.1 for testing purpose.

# Disable TLS certificate validation. If set to true, any certificate # presented by the server for any host name will be accepted # Deprecation: This option is now subsumed by the tls_config below. To # disable cert validation use: # tls_config { # disable_cert_validation: true # }

# TLS config

# Proxy URL, e.g. http://myproxy:3128

# HTTP proxy connect headers. These headers are passed on to the CONNECT # requests to the HTTP proxies. Note that CONNECT method is used to fetch # HTTPS URLs via HTTP proxies.

# User agent. Default user agent is Go's default user agent.

# Maximum idle connections to keep alive

# The maximum amount of redirects the HTTP client will follow. # To disable redirects, use max_redirects: 0.

# Add latency breakdown to probe results. This will add latency breakdown # by various stages of the request processing, e.g., DNS resolution, TCP # connection, TLS handshake, etc. You can select stages individually or # specify "ALL_STAGES" to get breakdown for all stages. # # Example: # latency_breakdown: [ ALL_STAGES ] # latency_breakdown: [ DNS_LATENCY, CONNECT_LATENCY, TLS_HANDSHAKE_LATENCY ]

# Parse HTTP response as additional metrics. If configured, Cloudprober # will try to extract metrics from HTTP response and export them along with # the default success/total/latency metrics.

# Interval between targets.

# Requests per probe. # Number of HTTP requests per probe. Requests are executed concurrently and # each HTTP re contributes to probe results. For example, if you run two # requests per probe, "total" counter will be incremented by 2.

# How long to wait between two requests to the same target. Only relevant # if requests_per_probe is also configured. # # This value should be less than (interval - timeout) / requests_per_probe. # This is to ensure that all requests are executed within one probe interval # and all of them get sufficient time. For example, if probe interval is 2s, # timeout is 1s, and requests_per_probe is 10, requests_interval_msec # should be less than 10ms.

# Packets per probe

# How long to wait between two packets to the same target

# Resolve targets after these many probes

# Ping payload size in bytes. It cannot be smaller than 8, number of bytes # required for the nanoseconds timestamp.

# Use datagram socket for ICMP. # This option enables unprivileged pings (that is, you don't require root # privilege to send ICMP packets). Note that most of the Linux distributions # don't allow unprivileged pings by default. To enable unprivileged pings on # some Linux distributions, you may need to run the following command: # # sudo sysctl -w net.ipv4.ping_group_range="0 <large valid group id>" # # net.ipv4.ping_group_range system setting takes two integers that specify # the group id range that is allowed to execute the unprivileged pings. Note # that the same setting (with ipv4 in the path) applies to IPv6 as well. # # Note: This option is not supported on Windows and is automatically # disabled there.

# Disable integrity checks. To detect data courruption in the network, we # craft the outgoing ICMP packet payload in a certain format and verify that # the reply payload matches the same format.

# Do not allow OS-level fragmentation, only works on Linux systems.

# Port for TCP requests. If not specfied, and port is provided by the # targets (e.g. kubernetes endpoint or service), that port is used.

# Whether to resolve the target before making the request. If set to false, # we hand over the target golang's net.Dial module, Otherwise, we resolve # the target first to an IP address and make a request using that. By # default we resolve first if it's a discovered resource, e.g., a k8s # endpoint.

# Interval between targets.

# Port to send UDP Ping to (UDP Echo). If running with the UDP server that # comes with cloudprober, it should be same as # ProberConfig.udp_echo_server_port.

# Number of sending side ports to use.

# message max to account for MTU.

# Payload size

# Changes the exported monitoring streams to be per port: # 1. Changes the streams names to total-per-port, success-per-port etc. # 2. Adds src_port and dst_port as stream labels. # Note that the field name is experimental and may change in the future.

# Whether to use all transmit ports per probe, per target. # Default is to probe each target once per probe and round-robin through the # source ports. # Setting this field to true changes the behavior to send traffic from all # ports to all targets in each probe. # For example, if num_tx_ports is set to 16, in every probe cycle, we'll send # 16 packets to every target (1 per tx port). # Note that setting this field to true will increase the probe traffic.

# maxTargets is the maximum number of targets supported by this probe type. # If there are more targets, they are pruned from the list to bring targets # list under maxTargets. A large number of targets has impact on resource # consumption.

# Port to listen.

# Number of packets sent in a single probe.

# Value can either be a static value or can be derived from target's labels. # To get value from target's labels, use target.labels.<target's label key> # as value.

# Whether to log metrics or not.

# Probe name. It should be unique across all probes.

# Interval between two probe runs in milliseconds. # Only one of "interval" and "inteval_msec" should be defined. # Default interval is 2s.

# Interval between two probe runs in string format, e.g. 10s. # Only one of "interval" and "inteval_msec" should be defined. # Default interval is 2s.

# Timeout for each probe in milliseconds # Only one of "timeout" and "timeout_msec" should be defined. # Default timeout is 1s.

# Timeout for each probe in string format, e.g. 10s. # Only one of "timeout" and "timeout_msec" should be defined. # Default timeout is 1s.

# Targets for the probe. Targets are required for all probes except # for external, user_defined, and extension probe types.

# Latency distribution. If specified, latency is stored as a distribution.

# Latency unit. Any string that's parseable by time.ParseDuration. # Valid values: "ns", "us" (or "µs"), "ms", "s", "m", "h".

# Latency metric name. You may want to change the latency metric name, if: # you're using latency_distribution for some probes, and regular metric for # other probes, and you want to differentiate between the two. # For example: # probe { # name: "web1_latency" # latency_distribution: {...} # latency_metric_name: "latency_dist" # ... # } # probe { # name: "app1" # ... # }

# Validators for this probe. Validators are run on the data returned by the # probe. See https://cloudprober.org/docs/how-to/validators/ for more info.

# Set the source IP to send packets from, either by providing an IP address # directly, or a network interface.

# How often to export stats. Probes usually run at a higher frequency (e.g. # every second); stats from individual probes are aggregated within # cloudprober until exported. In most cases, users don't need to change the # default. # # By default this field is set in the following way: # For all probes except UDP: # stats_export_interval=max(interval, 10s) # For UDP: # stats_export_interval=max(2*max(interval, timeout), 10s)

# Additional labels to add to the probe results. Label's value can either be # static or can be derived from target's labels. # # Example: # additional_label { # key: "app" # value: "@target.label.app@" # } # (More detailed example at: examples/additional_label/cloudprober.cfg)

# (Experimental) If set, test is inversed, i.e. we count it as success if # target doesn't respond. This is useful, for example, that your firewall is # working as expected. # # This is currently implemented only by PING and TCP probes. # Note: This field is currently experimental, and may change in future.

# Alerts configuration. If specified, cloudprober will generate alerts on # probe failures. You can specify multiple alerts. # Example: # alert { # name: "alert1" # condition {...} # notify { # pagerduty { ...} # } # } # alert { # name: "alert2" # notify { ... } # }

# Which machines this probe should run on. If defined, cloudprober will run # this probe only if machine's hostname matches this value. This is useful # for large deployments, where you may want to use the same prober config # everywhere but run this probe only on a subset of machines.

# Schedule for the probe. You can use a schedule to specify when a probe # should or should not run. This is useful for running probes only during # business hours. # # You can specify multiple schedules. Probe will not run if any of the # "DISABLE" schedules are active. If both "ENABLE" and "DISABLE" schedules # overlap, "DISABLE" takes precedence. # # For example, to disable a probe during weekends and on Tuesday between 7pm # and 8pm, e.g. for rollouts: # schdule { # type: DISABLE # start_weekday: FRIDAY # start_time: "20:00" # end_weekday: SUNDAY # end_time: "17:00" # timezone: "America/New_York" # } # schdule { # type: DISABLE # start_weekday: TUESDAY # start_time: "19:00" # end_weekday: TUESDAY # end_time: "20:00" # timezone: "America/New_York" # }

# Debug options. Currently only used to enable logging metrics.

# Period start weekday. If not specified, it defaults to EVERYDAY.

# Start time in 24 hour HH:MM format.

# Period end weekday. If not specified, it defaults to EVERYDAY.

# End time in 24 hour HH:MM format.

# Timezone in which the probe should run. If not specified, it defaults to # UTC. Example: "America/New_York"

# Azure account name and key. If you want to use managed identities, leave # account_key empty.

# Azure endpoint. Default is "https://<account>.blob.core.windows.net".

# OAuth2 configuration. If you want to use managed identities, leave this # field empty. See # https://cloudprober.org/docs/config/latest/oauth/#cloudprober_oauth_Config # for more details on oauth.Config.

# Serve test artifacts on Cloudprober's default webserver. This is # disabled by default for security reasons.

# Specify web server path to serve test artifacts on. # Default is "/artifacts/<probename>".

# Storage for test artifacts. Note that test artifacts are always # written to the workdir first, and uploaded to the storage backend in a # parallel goroutine. This is to make sure that uploads don't block the # main probe execution.

# Maximum age of artifacts in seconds.

# Cleanup interval in seconds. Default is 1 hour or max_age_sec, whichever # is smaller.

# If you want to use default credentials on GCE or GKE, leave this field # empty. See # https://cloudprober.org/docs/config/latest/oauth/#cloudprober_oauth_GoogleCredentials # for more details on oauth.GoogleCredentials.

# GCS endpoint.

# Cleanup options for local storage. Specifying cleanup options for local # enables automatic cleanup of old artifacts.

# Playwright test specs. These are passed to playwright as it is. This # field works in conjunction with test_dir -- test specs should be under # test directory. # # If test_spec is not specified, all test specs in the test directory are # executed, and since default test directory is config file's directory, # if you leave both the fields unspecified, all test specs co-located with # the config file are executed.

# Test directory. This is the directory where test specs are located. # Default test_dir is config file directory ("{{configDir}}").

# Workdir is path to the working directory. It should be writable. If not # specified, we try to create a temporary directory. All the output files # and reports are stored under <workdir>/output/. # If you need to be able access the output files, you should set this # field to a persistent location, e.g. a persistent volume, or configure # artifact options.

# Path to the playwright installation. We execute tests from this location. # If not specified, we'll use the value of environment variable # $PLAYWRIGHT_DIR, which is automatically set by the official cloudprober # playwright image (tag: "<version>-pw").

# NPX path. Default is to assume npx is in the PATH.

# Whether to enable screenshots for successful tests as well. # Note that screenshots are always enabled for failed tests, and you can # always save screenshots explicitly in the test spec.

# Traces are expensive and can slow down the test. We recommend to enable # this only when needed.

# By default, we export all test metrica as counters. You can change how # metrics are exported by setting the following options.

# Artifacts options.

# Cleanup options.

# Environment variables. These are passed/set before probing starts.

# Requests per probe. # Number of DNS requests per probe. Requests are executed concurrently and # each DNS request contributes to probe results. For example, if you run two # requests per probe, "total" counter will be incremented by 2.

# How long to wait between two requests to the same target. Only relevant # if requests_per_probe is also configured. # # This value should be less than (interval - timeout) / requests_per_probe. # This is to ensure that all requests are executed within one probe interval # and all of them get sufficient time. For example, if probe interval is 2s, # timeout is 1s, and requests_per_probe is 10, requests_interval_msec # should be less than 10ms.

# S3 endpoint. If not specified, default endpoint for the region is used.

# Domain to use when making DNS queries

# DNS Query Type

# Minimum number of answers expected. Default behavior is to return success # if DNS response status is NOERROR.

# Whether to resolve the target (target is DNS server here) before making # the request. If set to false, we hand over the target directly to the DNS # client. Otherwise, we resolve the target first to an IP address. By # default we resolve first if it's a discovered resource, e.g., a k8s # endpoint.

# DNS Query QueryClass

# Which DNS protocol is used for resolution.

# Requests per probe. # Number of DNS requests per probe. Requests are executed concurrently and # each DNS request contributes to probe results. For example, if you run two # requests per probe, "total" counter will be incremented by 2.

# How long to wait between two requests to the same target. Only relevant # if requests_per_probe is also configured. # # This value should be less than (interval - timeout) / requests_per_probe. # This is to ensure that all requests are executed within one probe interval # and all of them get sufficient time. For example, if probe interval is 2s, # timeout is 1s, and requests_per_probe is 10, requests_interval_msec # should be less than 10ms.

# Command. For ONCE probes, arguments are processed for the following field # substitutions: # @probe@ Name of the probe # @target.name@ or @target@ Hostname of the target # @target.port@ or @port@ Port of the target # @target.ip@ IP address associated with target # @address@ Resolved IP address of the target, in case of # discovered targets, same as @target.ip@. # @target.label.<x>@ Label x of the target # # For example, for target ig-us-central1-a, /tools/recreate_vm -vm @target@ # will get converted to: /tools/recreate_vm -vm ig-us-central1-a

# Command environment variables. These are passed on to the external probe # process as environment variables.

# Export output as metrics, where output is the output returned by the # external probe process, over stdout for ONCE probes, and through ProbeReply # for SERVER probes. Cloudprober expects variables to be in the following # format in the output: # var1 value1 (for example: total_errors 589)

# (Only applicable to ONCE mode). Disable streaming output metrics. By # default, external probe will export output metrics as they are available # on the stdout. If this option is set to true, output metrics will be # exported only after the probe has completed. # New in version 0.13.4. This was true by default in previous versions.

# Protoset contains descriptor source protos generated from the *.proto # files. You can use protoc to generate protoset files: # protoc --proto_path=. --descriptor_set_out=myservice.protoset \ # --include_imports my/custom/server/service.proto

# Note first 3 methods are valid only if descriptor source is not set.

# Request data (in JSON format) for the call_service_method request.

# Port for gRPC requests (Corresponding target field: port) # Default is 443, but if this field is not set and target has a port, either # discovered (e.g., k8s services, ingresses), or configured (e.g. endpoint, # file targets), we use target's port.

# Optional oauth config. For GOOGLE_DEFAULT_CREDENTIALS, use: # oauth_config: { bearer_token { gce_service_account: "default" } }

# If alts_config is provided, gRPC client uses ALTS for authentication and # encryption. For default alts configs, use: # alts_config: {}

# If TLSConfig is specified, it's used for authentication. # Note that only one of ALTSConfig and TLSConfig can be enabled at a time.

# if insecure_transport is set to true, TLS will not be used.

# Blob size for ECHO, READ, and WRITE methods.

# For HEALTH_CHECK, name of the service to health check.

# For HEALTH_CHECK, ignore status. By default, HEALTH_CHECK test passes # only if response-status is SERVING. Setting the following option makes # HEALTH_CHECK pass regardless of the response-status.

# Request definition for the GENERIC method.

# Number of connections to use. Default is 2 for ECHO, READ and WRITE # methods for backward compatibility. For HEALTH_CHECK and GENERIC, default # is 1.

# If connect_timeout is not specified, reuse probe timeout. Note that this # timeout will have an impact only if it is less than the probe timeout.

# URI scheme allows gRPC to use different resolvers # Example URI scheme: "google-c2p:///" # See https://github.com/grpc/grpc/blob/master/doc/naming.md for more details

# If provided, ALTS verifies that peer is using one of the given service # accounts.

# Handshaker service address. Default is to use the local metadata server. # For most of the ALTS use cases, default address should be okay.

# HTTP request scheme (Corresponding target label: "scheme"). If not set, we # use taget's 'scheme' label if present. # Note: protocol is deprecated, use scheme instead.

# Relative URL (Corresponding target label: "path"). We construct the final # URL like this: # <scheme>://<host>:<port>/<relative_url>. # # Note that the relative_url should start with a '/'.

# Port for HTTP requests (Corresponding target field: port) # Default is to use the scheme specific port, but if this field is not # set and discovered target has a port (e.g., k8s services, ingresses), # we use target's port.

# Whether to resolve the target before making the request. If set to true, # we resolve the target first to an IP address and make a request using # that while passing target name (or 'host' label if present) as Host # header. # # This behavior is automatic for discovered targets if they have an IP # address associated with them. Usually you don't need to worry about this # field and you can left it unspecified. We'll ty to do the right thing.

# Export response (body) count as a metric

# HTTP request method

# HTTP request headers # It is recommended to use "header" instead of "headers" for new configs. # header { # key: "Authorization" # value: "Bearer {{env "AUTH_TOKEN"}}" # }

# Request body. This field works similar to the curl's data flag. If there # are multiple "body" fields, we combine their values with a '&' in between. # # Also, we try to guess the content-type header based on the data: # 1) If data appears to be a valid json, we automatically set the # content-type header to "application/json". # 2) If the final data string appears to be a valid query string, we # set content-type to "application/x-www-form-urlencoded". Content type # header can still be overridden using the header field above. # Example: # body: "grant_type=client_credentials" # body: "scope=transferMoney" # body: "clientId=aweseomeClient" # body: "clientSecret=noSecret"

# Request body from file. This field is similar to the body field above, but # value is read from a file. # TODO(manugarg): We should consider providing a way to substitute environment # variables in the file.

# Enable HTTP keep-alive. If set to true, underlying connection is reused # for further probes. Default is to close the connection after every request.

# OAuth Config

# Disable HTTP2 # Golang HTTP client automatically enables HTTP/2 if server supports it. This # option disables that behavior to enforce HTTP/1.1 for testing purpose.

# Disable TLS certificate validation. If set to true, any certificate # presented by the server for any host name will be accepted # Deprecation: This option is now subsumed by the tls_config below. To # disable cert validation use: # tls_config { # disable_cert_validation: true # }

# TLS config

# Proxy URL, e.g. http://myproxy:3128

# HTTP proxy connect headers. These headers are passed on to the CONNECT # requests to the HTTP proxies. Note that CONNECT method is used to fetch # HTTPS URLs via HTTP proxies.

# User agent. Default user agent is Go's default user agent.

# Maximum idle connections to keep alive

# The maximum amount of redirects the HTTP client will follow. # To disable redirects, use max_redirects: 0.

# Add latency breakdown to probe results. This will add latency breakdown # by various stages of the request processing, e.g., DNS resolution, TCP # connection, TLS handshake, etc. You can select stages individually or # specify "ALL_STAGES" to get breakdown for all stages. # # Example: # latency_breakdown: [ ALL_STAGES ] # latency_breakdown: [ DNS_LATENCY, CONNECT_LATENCY, TLS_HANDSHAKE_LATENCY ]

# Parse HTTP response as additional metrics. If configured, Cloudprober # will try to extract metrics from HTTP response and export them along with # the default success/total/latency metrics.

# Interval between targets.

# Requests per probe. # Number of HTTP requests per probe. Requests are executed concurrently and # each HTTP re contributes to probe results. For example, if you run two # requests per probe, "total" counter will be incremented by 2.

# How long to wait between two requests to the same target. Only relevant # if requests_per_probe is also configured. # # This value should be less than (interval - timeout) / requests_per_probe. # This is to ensure that all requests are executed within one probe interval # and all of them get sufficient time. For example, if probe interval is 2s, # timeout is 1s, and requests_per_probe is 10, requests_interval_msec # should be less than 10ms.

# Packets per probe

# How long to wait between two packets to the same target

# Resolve targets after these many probes

# Ping payload size in bytes. It cannot be smaller than 8, number of bytes # required for the nanoseconds timestamp.

# Use datagram socket for ICMP. # This option enables unprivileged pings (that is, you don't require root # privilege to send ICMP packets). Note that most of the Linux distributions # don't allow unprivileged pings by default. To enable unprivileged pings on # some Linux distributions, you may need to run the following command: # # sudo sysctl -w net.ipv4.ping_group_range="0 <large valid group id>" # # net.ipv4.ping_group_range system setting takes two integers that specify # the group id range that is allowed to execute the unprivileged pings. Note # that the same setting (with ipv4 in the path) applies to IPv6 as well. # # Note: This option is not supported on Windows and is automatically # disabled there.

# Disable integrity checks. To detect data courruption in the network, we # craft the outgoing ICMP packet payload in a certain format and verify that # the reply payload matches the same format.

# Do not allow OS-level fragmentation, only works on Linux systems.

# Port for TCP requests. If not specfied, and port is provided by the # targets (e.g. kubernetes endpoint or service), that port is used.

# Whether to resolve the target before making the request. If set to false, # we hand over the target golang's net.Dial module, Otherwise, we resolve # the target first to an IP address and make a request using that. By # default we resolve first if it's a discovered resource, e.g., a k8s # endpoint.

# Interval between targets.

# Port to send UDP Ping to (UDP Echo). If running with the UDP server that # comes with cloudprober, it should be same as # ProberConfig.udp_echo_server_port.

# Number of sending side ports to use.

# message max to account for MTU.

# Payload size

# Changes the exported monitoring streams to be per port: # 1. Changes the streams names to total-per-port, success-per-port etc. # 2. Adds src_port and dst_port as stream labels. # Note that the field name is experimental and may change in the future.

# Whether to use all transmit ports per probe, per target. # Default is to probe each target once per probe and round-robin through the # source ports. # Setting this field to true changes the behavior to send traffic from all # ports to all targets in each probe. # For example, if num_tx_ports is set to 16, in every probe cycle, we'll send # 16 packets to every target (1 per tx port). # Note that setting this field to true will increase the probe traffic.

# maxTargets is the maximum number of targets supported by this probe type. # If there are more targets, they are pruned from the list to bring targets # list under maxTargets. A large number of targets has impact on resource # consumption.

# Port to listen.

# Number of packets sent in a single probe.