More flexibility to control server start #25790

dmatej · 2025-11-15T04:59:25Z

New parameters

User has control what will be considered as the "server successfully started" state.
- pid file
- process state
- any of admin ports
- all of custom endpoints (might be useful when we can access some public endpoint and we would need to block some script until the endpoint is publicly visible; I doubt if I would ever use it, but ...)
printServerOutput - explicitly ask to print the process output in the report. If not set, it is printed just if the terse was not set to true and if the start failed.

Logging

If terse not set, all start and restart commands now print some basic info about the server
- Can be extended later
When Restart Domain and Restart Local Instance now fail, they print the output of the failed start of the server. The start of the server already knows if it is restarting or just starting, so it also knows if there's some parent able to read the output. If there is no such parent, it will set the log handler and the local command then can read it.

Fixed Bugs

Admin endpoints can be a list, but we used just first item. At least one should be accessible.
Race conditions between processes
- Not sure if I can remember well and describe them well
- TBD, I have to read diff

- Added reporting important informations - More useful error management - --terse limits the amount of printed information - The fact that the process is alive, doesn't always mean it will be alive one second later. Check admin port too, by default. This will be configurable later. - User can configure "definition" of the started server - When process dies before the command, it is always an error - Added explicit option to print the process output - StartServerShutdownHook logging - it is kept, but the startup logging done by the AsadminMain is supported also on Windows Signed-off-by: David Matějček <[email protected]>

- It declared the debug port free, but immediately 300 ms later it was blocked. Signed-off-by: David Matějček <[email protected]>

Signed-off-by: David Matějček <[email protected]>

- We have issues on Windows on GHA with this. - Maybe local restarts should be done as stop-start on Windows? Respawning is not recommended and on windows it seems really risky Signed-off-by: David Matějček <[email protected]>

dmatej · 2025-11-26T22:41:04Z

appserver/itest-tools/src/main/java/org/glassfish/main/itest/tools/asadmin/Asadmin.java

    private AsadminResult exec(final Integer timeout, final boolean detachedAndTerse, final String... args) {
        final List<String> parameters = Arrays.asList(args);
-        LOG.log(TRACE, "exec(timeout={0}, detached={1}, args={2})", timeout, detachedAndTerse, parameters);
+        LOG.log(INFO, "exec(timeout={0}, detached={1}, args={2})", timeout, detachedAndTerse, parameters);


Not seeing this when I had issues on GHA was very confusing.

dmatej · 2025-11-26T22:42:19Z

...er/orb/orb-connector/src/main/java/org/glassfish/enterprise/iiop/api/GlassFishORBHelper.java


    @PostConstruct
    public void postConstruct() {
+        // WARN: Neither PreDestroy annotation nor interface worked!


PreDestroy annotation doesn't work here.
PreDestrou interface doesn't work too.
This was here on 2024, in one change I removed it, I was wrong.

dmatej · 2025-11-26T22:43:03Z

...er/orb/orb-connector/src/main/java/org/glassfish/enterprise/iiop/api/GlassFishORBHelper.java

+            // Still, threads already working with the instance will have it unstable.
+            final ORB destroyedOrb = orb;
            orb = null;
+            // FIXME: com.sun.corba.ee.impl.transport.AcceptorImpl.getAcceptedSocket(AcceptorImpl.java:127)


#25804 will target this.

dmatej · 2025-11-26T22:44:56Z

...rver/orb/orb-iiop/src/main/java/org/glassfish/enterprise/iiop/impl/IIOPSSLSocketFactory.java

-
-         serverSocket.bind(inetSocketAddress);
-         return serverSocket;
+        final Action<ServerSocket> action = () -> {


Desperate action, I will recheck in #25804 if it helps anything. When GF + Corba will reliably close all open ports on shutdown, this should be removed.

dmatej · 2025-11-26T22:49:16Z

...erver/tests/admin/tests/src/test/java/org/glassfish/main/admin/test/AsadminLoggingITest.java

-        assertThat(result, asadminOK());
+    public static void fillUpServerLog() throws Exception {
+        if (OS.isWindowsForSure()) {
+            // For some reason windows can collide on debug port.


I don't experience blocked --debug ports on Linux in restart-domain command, but GHA on Windows does it. Seems like it doesn't close debug port when one process spawns another - the new then tries to open the same port, could reuse it, which is not a good idea, but ...
I think we had this problem on Solaris many years ago too.

dmatej · 2025-11-26T22:50:38Z

appserver/tests/appserv-tests/devtests/ejb/ports.java

@@ -0,0 +1,76 @@
+/*


I tried to use AI to generate free ports, but finally this was easier.

dmatej force-pushed the restarts branch from 0dc1bd2 to 5120c6a Compare November 17, 2025 01:35

dmatej changed the title ~~Server starts~~ More flexibility to control server start Nov 18, 2025

dmatej requested a review from a team November 18, 2025 00:54

dmatej added this to the 7.1.0 milestone Nov 18, 2025

dmatej linked an issue Nov 18, 2025 that may be closed by this pull request

The start-domain command randomly times out #25615

Closed

arjantijms approved these changes Nov 18, 2025

View reviewed changes

dmatej force-pushed the restarts branch 14 times, most recently from ad8c121 to a50c732 Compare November 21, 2025 13:08

This was referenced Nov 21, 2025

Set of smaller improvements #25797

Merged

AutoDisposableGlassFish failure: no CommandRunner implementation available #25705

Open

dmatej force-pushed the restarts branch from a50c732 to 9c4ab48 Compare November 22, 2025 01:09

dmatej mentioned this pull request Nov 22, 2025

GFLauncher revisited #25801

Merged

dmatej force-pushed the restarts branch 2 times, most recently from 72f5d51 to a8cbf02 Compare November 22, 2025 21:41

dmatej mentioned this pull request Nov 23, 2025

ListDomainsCommand now lists all admin endpoints, cleanup #25802

Merged

dmatej force-pushed the restarts branch 2 times, most recently from 3918e8a to a9d06c9 Compare November 23, 2025 12:29

dmatej force-pushed the restarts branch from a9d06c9 to 0025518 Compare November 23, 2025 19:23

dmatej mentioned this pull request Nov 24, 2025

Corba refactoring #25804

Draft

dmatej added 2 commits November 24, 2025 12:07

The waitWhileListening did not work

8e196fe

- It declared the debug port free, but immediately 300 ms later it was blocked. Signed-off-by: David Matějček <[email protected]>

Fixed ORB cleanup and using ports after fast restarts (reuse)

bc8798d

Signed-off-by: David Matějček <[email protected]>

dmatej force-pushed the restarts branch from 6d3507b to f89b88c Compare November 25, 2025 10:32

Added port checking to ORB

2f2ceac

Signed-off-by: David Matějček <[email protected]>

dmatej force-pushed the restarts branch from f89b88c to 2f2ceac Compare November 25, 2025 11:32

Improved port checking, reduced logging

d62136f

Signed-off-by: David Matějček <[email protected]>

dmatej force-pushed the restarts branch from d813697 to d62136f Compare November 25, 2025 21:05

dmatej added 4 commits November 26, 2025 00:36

Use random ports rather than hard coded

cf25c6d

Signed-off-by: David Matějček <[email protected]>

Retry binding to avoid zombies + postpone shutdown

d615f2e

Signed-off-by: David Matějček <[email protected]>

Better random port generator written in java

5234fa5

Signed-off-by: David Matějček <[email protected]>

Fixed reporting when the start failed.

f103aa2

Signed-off-by: David Matějček <[email protected]>

dmatej force-pushed the restarts branch from 34118fe to 7d79d9b Compare November 26, 2025 19:11

dmatej force-pushed the restarts branch from 7d79d9b to 01101fb Compare November 26, 2025 22:30

dmatej marked this pull request as ready for review November 26, 2025 22:40

dmatej commented Nov 26, 2025

View reviewed changes

arjantijms merged commit 342086c into eclipse-ee4j:master Nov 27, 2025
6 checks passed

dmatej added the enhancement New feature or request label Nov 27, 2025

dmatej deleted the restarts branch November 27, 2025 13:51

dmatej mentioned this pull request Nov 28, 2025

Documentation for --server-output, plus fixes after testing #25807

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

More flexibility to control server start #25790

More flexibility to control server start #25790

Uh oh!

dmatej commented Nov 15, 2025 •

edited

Loading

Uh oh!

dmatej Nov 26, 2025

Uh oh!

dmatej Nov 26, 2025

Uh oh!

dmatej Nov 26, 2025

Uh oh!

dmatej Nov 26, 2025

Uh oh!

dmatej Nov 26, 2025

Uh oh!

dmatej Nov 26, 2025

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

More flexibility to control server start #25790

More flexibility to control server start #25790

Uh oh!

Conversation

dmatej commented Nov 15, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

New parameters

Logging

Fixed Bugs

Uh oh!

dmatej Nov 26, 2025

Choose a reason for hiding this comment

Uh oh!

dmatej Nov 26, 2025

Choose a reason for hiding this comment

Uh oh!

dmatej Nov 26, 2025

Choose a reason for hiding this comment

Uh oh!

dmatej Nov 26, 2025

Choose a reason for hiding this comment

Uh oh!

dmatej Nov 26, 2025

Choose a reason for hiding this comment

Uh oh!

dmatej Nov 26, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

dmatej commented Nov 15, 2025 •

edited

Loading