@Rocky - an application connecting from a dedicated service account to starrocks 3.5.20 that runs in shared-data mode is getting periodic error "error connecting: Timeout expired. The timeout period elapsed prior to obtaining a connection from the pool." What can be done to resolve this issue?

The error message "Timeout expired. The timeout period elapsed prior to obtaining a connection from the pool" is a client-side error (common in .NET, JDBC, and other connection poolers). It indicates that the application's internal connection pool is exhausted, and no connection became available within the configured timeout period. While the error originates in your application, it is often triggered by server-side bottlenecks or configuration limits in StarRocks. To resolve this in StarRocks 3.5.20 (shared-data mode), follow these steps: 1. Check User-Specific Connection Limits Since you mentioned a dedicated service account, check if there is a per-user connection limit. StarRocks allows restricting the number of concurrent connections for a specific user. * Check limit:

SHOW PROPERTY FOR 'your_service_account';

* Increase limit: If

max_user_connections

is set low, increase it:

sql
    ALTER USER 'your_service_account' SET PROPERTIES ("max_user_connections" = "2000");

2. Verify Global Connection Limits (

qe_max_connection

) StarRocks FEs have a global limit on concurrent MySQL connections. In version 3.5.x, the default is 4096. * Check current connections: Run

SHOW PROCESSLIST;

or check the

starrocks_fe_query_resource_group_running_queries

metric. * Check FE config: Run

ADMIN SHOW CONFIG LIKE 'qe_max_connection';

* Solution: If you are hitting this limit, increase it in `fe.conf`:

properties
    qe_max_connection = 8192

(Note: This requires an FE restart.) 3. Enable NIO for MySQL Service In high-concurrency environments, the default blocking I/O for MySQL connections can cause bottlenecks in the FE. Enabling Netty-based Non-blocking I/O (NIO) improves the FE's ability to handle many concurrent connections. * Edit `fe.conf`:

properties
    mysql_service_nio_enabled = true

(Note: This requires an FE restart.) 4. Optimize Shared-Data FE Metadata Pressure In shared-data mode, the FE performs extra work for tablet metadata management and compaction scheduling. If the FE is under heavy load (high CPU or JVM GC), connection handshakes may time out. * Monitor FE JVM: Check for frequent "Stop the World" GC pauses. * Publish Version Threads: If you have high-frequency ingestion, increase the threads available for version publishing in `fe.conf`:

properties
    lake_publish_version_max_threads = 1024

5. Application-Side Troubleshooting If StarRocks appears healthy, the issue likely resides in the application's connection management: * Connection Leaks: Ensure the application always closes or returns connections to the pool in a

finally

block. * Increase Pool Size: Increase the

Max Pool Size

in your application’s connection string (e.g.,

Max Pool Size=200;

). * Increase Timeout: Increase the

Connection Timeout

or

Connect Timeout

in the connection string to allow more time for a connection to become available during spikes. Summary Checklist 1. Run

SHOW PROCESSLIST

to see if connections are stuck in a

Sleep

state (leak) or running slow queries. 2. Increase

max_user_connections

for the service account. 3. Set

mysql_service_nio_enabled = true

in

fe.conf

. 4. Increase the application's

Max Pool Size

. References • faq/shared_data_faq.md • quick_start/shared-data.md