Monitor Microsoft Fabric long-running operations — fabric_operation

Check, wait for, and retrieve the result of a Fabric operation that continues after its initiating request returns. Pass the operation handle returned by a 'fabricQueryR' function when possible. To resume work later, save the complete Location URL returned by Fabric. A bare operation ID can reconstruct only the core /operations/{id} route, not workload-scoped routes

Usage

fabric_operation_status(
  operation,
  tenant_id = Sys.getenv("FABRICQUERYR_TENANT_ID"),
  client_id = Sys.getenv("FABRICQUERYR_CLIENT_ID", unset =
    "04b07795-8ddb-461a-bbee-02f9e1bf7b46"),
  token = NULL,
  auth_args = list(),
  api_base = .fabric_api_base,
  respect_retry_after = TRUE,
  .sleep = Sys.sleep,
  .now = Sys.time
)

fabric_operation_wait(
  operation,
  poll_interval = NULL,
  timeout = 300,
  error_on_failure = TRUE,
  tenant_id = Sys.getenv("FABRICQUERYR_TENANT_ID"),
  client_id = Sys.getenv("FABRICQUERYR_CLIENT_ID", unset =
    "04b07795-8ddb-461a-bbee-02f9e1bf7b46"),
  token = NULL,
  auth_args = list(),
  api_base = .fabric_api_base,
  .sleep = Sys.sleep,
  .now = Sys.time
)

fabric_operation_result(
  operation,
  wait = TRUE,
  poll_interval = NULL,
  timeout = 300,
  tenant_id = Sys.getenv("FABRICQUERYR_TENANT_ID"),
  client_id = Sys.getenv("FABRICQUERYR_CLIENT_ID", unset =
    "04b07795-8ddb-461a-bbee-02f9e1bf7b46"),
  token = NULL,
  auth_args = list(),
  api_base = .fabric_api_base,
  .sleep = Sys.sleep,
  .now = Sys.time
)

Arguments

operation: A fabric_operation handle, Fabric operation GUID, or operation state/result URL returned in a Location header
tenant_id: Entra tenant ID. Defaults to FABRICQUERYR_TENANT_ID
client_id: Entra application ID. Defaults to FABRICQUERYR_CLIENT_ID, then the Azure CLI application ID
token: Optional access token or token-provider function. Leave NULL to let 'fabricQueryR' use its normal sign-in flow. A fabric_operation handle reuses its stored credential unless authentication arguments are supplied explicitly
auth_args: Additional sign-in options passed to AzureAuth::get_azure_token() when no token source is supplied
api_base: Fabric REST API base URL. Most users should keep the default
respect_retry_after: Whether to wait until Fabric's recommended next status-check time before making the request
.sleep, .now: Internal hooks for deterministic tests
poll_interval: Minimum seconds between status requests. NULL honors Fabric's Retry-After value and otherwise uses a two-second fallback
timeout: Positive maximum total seconds to wait, including status requests
error_on_failure: Whether a failed operation should raise a fabric_operation_failed condition. Set to FALSE to inspect the returned failed state directly
wait: Whether to wait for a running operation. When FALSE, one state request is made immediately without honoring a stored future polling hint; a non-terminal operation raises fabric_operation_not_ready.

Value

fabric_operation_status() and fabric_operation_wait() return a fabric_operation_state record. fabric_operation_result() returns a fabric_operation_result with value, content_type, empty, HTTP and request identifiers, and the reusable operation handle. JSON results are decoded as lists, binary results are raw vectors, and empty results have a NULL value

Typical workflow

A package function that starts asynchronous work may return a fabric_operation handle. Use fabric_operation_wait() to wait for it to finish and fabric_operation_result() to retrieve its output. Result retrieval waits by default, so it is enough for the common case

If the R process restarts, save the service-provided location and pass it with fresh authentication arguments. A bare ID is sufficient only for core operations; after success, core operations are completed by reading the documented /operations/{id}/result resource

Results and failures

fabric_operation_status() preserves Fabric's status, progress, timestamps, request identifiers, and structured error. Status values added by Fabric in the future remain inspectable, but fabric_operation_wait() stops with a typed error instead of polling an unfamiliar value indefinitely. The documented Undefined, NotStarted, and Running values remain pending

Some workload APIs, including Lakehouse table loading, expose completion in their state response and do not provide a separate /result resource. For those operations, fabric_operation_result() returns the terminal state payload as its value

A failed operation raises fabric_operation_failed by default. A timeout raises fabric_operation_timeout; neither condition repeats the request that originally started the operation

Regional operation endpoints

Fabric can return a Location on a regional *.analysis.windows.net cluster. 'fabricQueryR' recognizes those Microsoft endpoints and automatically uses the Power BI token audience they require. Normal automatic sign-in or an audience-aware token-provider function handles both audiences. A single static Fabric bearer token cannot authenticate a regional operation URL

References

Get operation state

Get operation result

Regional Fabric LRO authentication example

Examples

if (FALSE) { # \dontrun{
# Discover a Lakehouse and a CSV file that Fabric can load as a table
workspace <- fabric_workspaces()[[1L]]
lakehouse <- fabric_lakehouses(workspace)[[1L]]
files <- fabric_onelake_list(
  workspace,
  lakehouse,
  path = "Files/incoming"
)
csv_file <- files[grepl("[.]csv$", files$path), ][1L, ]

# The load call returns the long-running operation handle used below
operation <- fabric_lakehouse_load_table(
  lakehouse,
  table = "orders_imported",
  path = csv_file$path[[1L]],
  format = "Csv",
  header = TRUE
)

# Check once, wait for completion, then retrieve the operation result
state <- fabric_operation_status(operation)
completed <- fabric_operation_wait(state$operation, timeout = 900)
result <- fabric_operation_result(completed$operation)
result$value
} # }