@Rocky I am trying to load one small json file that exists in GCS (81.6 KB) I have permissions to it from my starrocks account. I am using compute nodes and when running show compute nodes I see they are live and OK. There are no other LOADs in progress, only mine exists and is QUEUEING The LOAD is being created but then stuck on state:QUEUEING and doesn't start. how can i debug it? what can i do? this is the code:

LOAD LABEL org_id_to_name_load_test_label_2
(
    DATA INFILE("<gs://data-ads-core-forge-prd/bigquery_exports/unity-data-ads-core-prd/druid_lookups/org_id_to_name/*>")
    INTO TABLE org_id_to_name_load_test
    FORMAT AS "json"
 )
 WITH BROKER
 (
 
    "gcp.gcs.use_compute_engine_service_account" = "true"
 )
PROPERTIES
(
    "timeout" = "300"
);

Being stuck in the

QUEUEING

state for a Broker Load usually indicates that the StarRocks Frontend (FE) has received the request but hasn't yet allocated a thread or a slot to begin the execution. Since you mentioned there are no other loads in progress and your nodes are healthy, here is a step-by-step guide to debug and resolve this: 1. Check FE Concurrency Limits StarRocks has internal limits on how many Broker Load jobs can run concurrently across the cluster. If these limits are reached (even by "ghost" jobs or jobs in other databases), new jobs will sit in the

QUEUEING

state. Run the following command to check your current configuration:

sql
ADMIN SHOW FRONTEND CONFIG LIKE '%load_job_concurrency%';

*

max_broker_load_job_concurrency

: The default is usually 5. If there are other loads (perhaps in different databases or routine loads) or if this was accidentally set to 0, your job will stay queued. *

max_running_txn_num_per_db

: Check if there are many open transactions in your specific database. 2. Check for Hidden Loads or Routine Loads Even if you don't see "Broker Loads" running, other asynchronous processes might be consuming the scheduler's capacity. * Check for any Routine Loads that might be running:

SHOW ROUTINE LOAD;

* Check for any Insert Into tasks or Schema Changes that might be holding locks:

SHOW ALTER TABLE COLUMN;

* Verify all loads across the cluster:

SHOW LOAD WHERE state = 'LOADING' OR state = 'QUEUEING';

3. Verify FE Leader Status Broker Load scheduling is handled by the FE Leader. If you are connected to a Follower FE that is having trouble communicating with the Leader, scheduling might hang. * Run

SHOW FRONTENDS;

and ensure there is one node where

IsMaster

(or

IsLeader

) is

true

and

Alive

is

true

. 4. Inspect FE Logs (The most definitive method) Since the job is stuck in the FE's scheduler, the reason is almost always logged in

fe.log

. 1. Find your

fe.log

file on the FE Leader node. 2. Search for your

LABEL

or the

JobId

(which you can get from

SHOW LOAD

). 3. Look for messages like: *

waiting for next schedule...

*

load job concurrency reach the limit

*

failed to create etl tasks

5. Check Resource Group Constraints If you are using Workload Management (Resource Groups), your user or the load task might be assigned to a group with very low concurrency or no available CPU/Memory. _ Check if you have resource groups configured:

SELECT _ FROM information_schema.resource_groups;

6. Common Fixes to Try If the cluster appears idle, you can try to "nudge" the scheduler by slightly increasing the concurrency limit dynamically:

sql
ADMIN SET FRONTEND CONFIG ("max_broker_load_job_concurrency" = "10");

Potential Syntax Issue (Minor) While your syntax is generally correct for GCS, ensure that your FE has internet access (or VPC access) to GCS to perform the initial file listing (the "Plan" phase). If the FE cannot list the files in

gs://.../*

, it can sometimes hang before transitioning to

LOADING

. However, since you are using

WITH BROKER

without a specific broker name, StarRocks is using its native cloud-loading capability (which is correct for modern versions). Ensure that your StarRocks version is at least 2.4+, as earlier versions required an actual Broker process for GCS. Summary of action: 1. Run

SHOW LOAD

and grab the

JobId

. 2. Check

fe.log

on the Leader FE for that

JobId

. 3. Check

ADMIN SHOW FRONTEND CONFIG LIKE 'max_broker_load_job_concurrency';

. References • loading/gcs.md • sql-reference/sql-statements/loading_unloading/BROKER_LOAD.md • loading/BrokerLoad.md