GitHub - microsoft/dcp: Developer Control Plane API server and CLI.

DCP_BIN_PATH and DCP_EXTENSIONS_PATH These variables point to, respectively, the DCP root binary/configuration directory, and the DCP extensions directory. DCP root directory contains the main DCP CLI/API server binary and the default access configuration (kubeconfig) file. The extensions directory contains the DCP controller process binary and other extensions, if present.
By default, DCP assumes the root directory to be ${HOME}/.dcp. .NET Aspire tooling is using these environment variables to instruct DCP to use locations and binaries inside Aspire.Hosting.Orchestration workload instead. DEBUG_SESSION_PORT, DEBUG_SESSION_TOKEN, and DEBUG_SESSION_SERVER_CERTIFICATE These are variables that configure the endpoint for running Executables via a developer IDE/under debugger. For more information see IDE execution specification. DCP_SESSION_FOLDER This variable is used for isolating multiple DCP instances running concurrently on the same machine. If set (to a valid filesystem folder), DCP process(es) will create files related to their execution in this folder: the access configuration file (kubeconfig), captured Executable/Container logs, etc. DCP_LOG_SOCKET If set to a Unix domain socket, DCP will write its execution logs to that socket instead of writing them to standard error stream (stderr). This allows programs that launch DCP to capture its output even if DCP is running in --detach mode.
The --detach mode causes DCP to fork itself and break the parent-child relationship (and lifetime dependency) from the process that launched it, but the side effect of doing so is that the parent process loses ability to monitor DCP standard output and standard error streamd. DCP_LOG_SESSION_ID If set, DCP will prepend this value to all diagnostics log names. If unset, a session ID will be calculated. The value is propagated to all child DCP processes. DCP_DIAGNOSTICS_LOG_LEVEL If set, enabled DCP diagnostic logging.
Can be set to error, info, or debug; for troubleshooting debug is recommended, although it results in the most verbose output. DCP_DIAGNOSTICS_LOG_FOLDER If set to a valid filesystem folder, DCP will place the diagnostic logging files there. Otherwise (if enabled) they are written to the default temporary files folder. DCP_LOG_FILE_NAME_SUFFIX Suffix to append to the log file name (defaults to process ID if not set). DCP_LOGGING_CONTEXT If set, the value of this variable will be written to the log file as one of the first log messages (as verbose, "info" type of message). DCP_PRESERVE_EXECUTABLE_LOGS If set (to "true", "yes", or "1"), the logs from Executables will not be deleted when DCP shuts down. This can be useful to capture results of test runs that use DCP as the workload orchestrator. DCP_RESOURCE_WATCH_TIMEOUT_SECONDS A timeout for resource watch requests, in seconds. Watch requests will time out shortly after the specified value, to avoid the "thundering herd" problem. Useful for testing watch retry logic. DCP_IDE_REQUEST_TIMEOUT_SECONDS, DCP_IDE_NOTIFICATION_TIMEOUT_SECONDS, and DCP_IDE_NOTIFICATION_KEEPALIVE_SECONDS Timeouts for, respectively: requests to IDE run session endpoint (defaults to 120 seconds), request to IDE notification WebSocket endpoint (defaults to 20 seconds), and the time to interpret the lack of response to WebSocket ping message on the IDE notification connection is interpreted as connection failue (defaults to 5 seconds).
The IDE notification keepalive value, if present, must be smaller than the IDE notification request timeout value.
A value of zero for IDE notification keepalive turns off the keep-alives (ping-pong messages) for the IDE notification connection. DCP_INSTANCE_ID_PREFIX A prefix for the automatically generated, random DCP instance ID, which is attached to every communication with the IDE when running Executables via IDE. For more information see IDE execution specification. DCP_MOST_RECENT_PORT_LIFETIME, DCP_PORT_AVAILABILITY_CHECK_TIMEOUT, DCP_PORT_ALLOCATION_CHECK_TIMEOUT, DCP_PORT_ALLOCATIONS_PER_ROUND, and DCP_PORT_ALLOCATION_ROUND_DELAY These are parameters that govern the algorithm DCP is using to ensure that dynamically allocated ports are unique for the whole machine, even if multiple DCP instances are running.

DCP_MOST_RECENT_PORT_LIFETIME is the time a recently allocated network port is considered unavailable. Used during automatic port allocation. Defaults to 2 minutes. Note that a particular port number will never be re-used if it is being actively used by a program running on the machine, even if that port has been auto-allocated earlier than "lifetime" ago.

DCP_PORT_AVAILABILITY_CHECK_TIMEOUT determines how long DCP will try to access the most recently used ports file when checking for port availability. Defaults to 50 milliseconds.

DCP_PORT_ALLOCATION_CHECK_TIMEOUT determines how long DCP will try to ensure that the allocated port is unique for the whole machine (per single port allocation). Defaults to 1 second.

DCP_PORT_ALLOCATIONS_PER_ROUND determines how many port allocation attempts a DCP instance will make while holding a lock on the most recently used ports file. Defaults to 5.

DCP_PORT_ALLOCATION_ROUND_DELAY tells DCP how long to wait (releasing the lock on the MRU ports file) before attempting to allocate a unique port again. Defaults to 100 ms.

The values of all these environment variables should be the same as Go time.Duration literals (except from DCP_PORT_ALLOCATIONS_PER_ROUND, which is a positive integer).

DCP_IP_VERSION_PREFERENCE Describes which version of the IP protocol will be preferred for communicating with DCP and for allocating service ports. Can be set to IPv4, v4 or 4 (to prefer IP protocol version 4) or IPv6, v6 or 6 (to prefer IP protocol version 6).

The preference determines which type of available IP addresses will be preferred when resolving localhost host name. localhost is the default pseudo-address that both Services and DCP itself bind to. The value of DCP_IP_VERSION_PREFERENCE environment variable is not used if a Service requests specific address, or if a Service uses an address allocation mode that is different from localhost.

DCP_SHUTDOWN_TIMEOUT_SECONDS Overrides the default DCP shutdown timeout (120 seconds). DCP will use this timeout to wait for all resources to be cleaned up before shutting down. DCP_SECURE_TOKEN Provides a predetermined bearer token for DCP to use rather than generating a randm one. If this is set, DCP will write a placeholder value to the kubeconfig file. Any client trying to connect will have to know this predetermined token and apply it to their config themselves. DCP_PERF_TRACE If set, instructs DCP to capture a performance trace during startup and/or shutdown. For more information see performance investigations page. DCP_DISABLE_PROCESS_CLEANUP_JOB On Windows, DCP will use a Win32 Job object to ensure that processes it launches are terminated when it shuts down. This alters standard process startup sequence and might not work on tightly locked-down machines, or might be flagged as "suspicious behavior" by antivirus software. If DCP cannot start processes and complains about "process cleanup job" errors, setting DCP_DISABLE_PROCESS_CLEANUP_JOB to 1 or true should help.