[
https://issues.apache.org/jira/browse/IGNITE-28329?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Ashish Mishra updated IGNITE-28329:
-----------------------------------
Description:
When an AI3 Example is executed repeatedly in a loop against a cluster running
in Docker, the example hangs indefinitely on a JDBC executeUpdate call (either
CREATE TABLE or a subsequent DDL statement) within the first 10 iterations. The
hang is permanent - no timeout fires and no exception is thrown. The issue is
not reproducible with ZIP distribution.
Environment
Client: OS Windows JDK 17
Server: Docker
Steps to Reproduce
Start an Ignite cluster using the Docker distribution (server running inside
Docker, client on Windows host)
Run SqlJdbcExampleRunner or any example with DDL and DML statements configured
to loop for 15 minutes (default loop interval)
Observe that within ~10 iterations, the example hangs permanently
You can use my attached reproducer as well
Expected Behavior
Each iteration of SqlJdbcExample completes successfully and the next iteration
begins. The example should run stably for the full configured duration without
hanging.
Actual Behavior
After a variable number of successful iterations (typically fewer than 10),
the example gets stuck on executeUpdate at SqlJdbcExample.java:55 and never
returns. The example runner logs the hang:
[3] Completed in 4423 ms.
Connecting to server...
[4] HUNG after 60016 ms.
The hung thread stack at the time of failure:
Thread: pool-1-thread-1 [WAITING]
at java.util.concurrent.CompletableFuture.join(CompletableFuture.java:2117)
at
org.apache.ignite.internal.jdbc.JdbcStatement.execute0(JdbcStatement.java:185)
at
org.apache.ignite.internal.jdbc.JdbcStatement.executeUpdate(JdbcStatement.java:198)
at
org.apache.ignite.example.sql.jdbc.SqlJdbcExample.main(SqlJdbcExample.java:55)
The CompletableFuture inside JdbcStatement.execute0 never completes. The TCP
client channel heartbeat thread remains alive and Netty NIO is polling
normally, confirming the connection to the cluster is intact. The hang is
server-side — the cluster receives (or drops) the request but never sends a
response.
From extended watchdog logs (stability test environment), the exact same stack
is observed frozen for 11+ consecutive minutes with zero state change.
was:
When an AI3 Example is executed repeatedly in a loop against a cluster running
in Docker, the example hangs indefinitely on a JDBC executeUpdate call (either
CREATE TABLE or a subsequent DDL statement) within the first 10 iterations. The
hang is permanent - no timeout fires and no exception is thrown. The issue is
not reproducible with ZIP distribution.
Environment
PropertyValueClient OSWindowsClient JDKJDK 17Server deploymentDocker
distributionBuild toolGradleReproduced onAI3, GG9
Steps to Reproduce
Start an Ignite cluster using the Docker distribution (server running inside
Docker, client on Windows host)
Run SqlJdbcExampleRunner or any example with DDL and DML statements configured
to loop for 15 minutes (default loop interval)
Observe that within ~10 iterations, the example hangs permanently
You can use my attached reproducer as well
Expected Behavior
Each iteration of SqlJdbcExample completes successfully and the next iteration
begins. The example should run stably for the full configured duration without
hanging.
Actual Behavior
After a variable number of successful iterations (typically fewer than 10), the
example gets stuck on executeUpdate at SqlJdbcExample.java:55 and never
returns. The example runner logs the hang:
[3] Completed in 4423 ms.
Connecting to server...
[4] HUNG after 60016 ms.
The hung thread stack at the time of failure:
Thread: pool-1-thread-1 [WAITING]
at java.util.concurrent.CompletableFuture.join(CompletableFuture.java:2117)
at
org.apache.ignite.internal.jdbc.JdbcStatement.execute0(JdbcStatement.java:185)
at
org.apache.ignite.internal.jdbc.JdbcStatement.executeUpdate(JdbcStatement.java:198)
at
org.apache.ignite.example.sql.jdbc.SqlJdbcExample.main(SqlJdbcExample.java:55)
The CompletableFuture inside JdbcStatement.execute0 never completes. The TCP
client channel heartbeat thread remains alive and Netty NIO is polling
normally, confirming the connection to the cluster is intact. The hang is
server-side — the cluster receives (or drops) the request but never sends a
response.
>From extended watchdog logs (stability test environment), the exact same stack
>is observed frozen for 11+ consecutive minutes with zero state change.
> AI3 examples hangs indefinitely on DDL statements during cyclic execution
> against Docker distribution
> -----------------------------------------------------------------------------------------------------
>
> Key: IGNITE-28329
> URL: https://issues.apache.org/jira/browse/IGNITE-28329
> Project: Ignite
> Issue Type: Bug
> Components: examples ai3, jdbc ai3
> Reporter: Ashish Mishra
> Assignee: Ashish Mishra
> Priority: Critical
> Labels: ignite-3
> Fix For: 3.2
>
> Attachments: SqlJdbcExampleRunner.java
>
>
> When an AI3 Example is executed repeatedly in a loop against a cluster
> running in Docker, the example hangs indefinitely on a JDBC executeUpdate
> call (either CREATE TABLE or a subsequent DDL statement) within the first 10
> iterations. The hang is permanent - no timeout fires and no exception is
> thrown. The issue is not reproducible with ZIP distribution.
> Environment
> Client: OS Windows JDK 17
> Server: Docker
> Steps to Reproduce
> Start an Ignite cluster using the Docker distribution (server running inside
> Docker, client on Windows host)
> Run SqlJdbcExampleRunner or any example with DDL and DML statements
> configured to loop for 15 minutes (default loop interval)
> Observe that within ~10 iterations, the example hangs permanently
> You can use my attached reproducer as well
> Expected Behavior
> Each iteration of SqlJdbcExample completes successfully and the next
> iteration begins. The example should run stably for the full configured
> duration without hanging.
> Actual Behavior
> After a variable number of successful iterations (typically fewer than 10),
> the example gets stuck on executeUpdate at SqlJdbcExample.java:55 and never
> returns. The example runner logs the hang:
> [3] Completed in 4423 ms.
> Connecting to server...
> [4] HUNG after 60016 ms.
> The hung thread stack at the time of failure:
> Thread: pool-1-thread-1 [WAITING]
> at java.util.concurrent.CompletableFuture.join(CompletableFuture.java:2117)
> at
> org.apache.ignite.internal.jdbc.JdbcStatement.execute0(JdbcStatement.java:185)
> at
> org.apache.ignite.internal.jdbc.JdbcStatement.executeUpdate(JdbcStatement.java:198)
> at
> org.apache.ignite.example.sql.jdbc.SqlJdbcExample.main(SqlJdbcExample.java:55)
> The CompletableFuture inside JdbcStatement.execute0 never completes. The TCP
> client channel heartbeat thread remains alive and Netty NIO is polling
> normally, confirming the connection to the cluster is intact. The hang is
> server-side — the cluster receives (or drops) the request but never sends a
> response.
> From extended watchdog logs (stability test environment), the exact same
> stack is observed frozen for 11+ consecutive minutes with zero state change.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
