… match (#15309)

Two cluster bugs tracked by #15032 (supersedes the abandoned #14951 and
addresses the review feedback there):

1. ServerMemberManager.hasMember(String): the IP-only fallback used a substring
   match (StringUtils.contains), so "192.168.1.10" wrongly matched a member
   "192.168.1.100:8848", causing false-positive membership and split-brain
   alerts. Compare the member IP exactly (member.getIp().equals(address)),
   which is also IPv6-safe (no ":" splitting).

2. TcpHealthCheckProcessor (v2) PostProcessor: the successful connect path did
   not close the channel, and the ConnectException (connection-refused) path
   did not either. On failure finishCheck() removes the beat from keyMap, so the
   next round can no longer close the orphaned channel, leaking a file
   descriptor on every connection-refused check. Cancel the key and close the
   channel on both paths.

Add focused regression tests: hasMember prefix/substring collisions in
ServerMemberManagerTest, and channel-closed-on-success / connection-refused in
a new TcpHealthCheckProcessorTest (real loopback sockets + real Selector).

Assisted-by: Claude Code

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
(cherry picked from commit fd5114f)

(cherry picked from commit ea17ef8)

…ce detail (#15348)

* [ISSUE #15296] Fix instance toggle not updating service detail UI

The console instance list endpoint serves a cached view that is refreshed asynchronously after an accepted write (naming ServiceStorage cache, push task delay defaults to 500ms), so refetching immediately after updateInstance rendered the old enabled state and the toggle appeared to need two clicks.

Patch the confirmed values into local state instead of refetching, the same approach the legacy console InstanceTable uses. Apply the same handling to the edit-instance dialog, which refetched the service detail and every cluster instance list through the same stale cache.

Assisted-by: Claude Code

* [ISSUE #15296] Fix deleted instance reappearing in service detail

The instance list endpoint serves a cached ServiceInfo that is refreshed asynchronously after an accepted write, so refetching right after deleteInstance can render the deleted row again. Remove the row from local state instead, the same confirmed-local-update approach already used for the online/offline toggle and the edit-instance dialog.

The previous-page navigation (deleting the last row of a page > 1) still has to fetch since that page is not held locally; the local removal afterwards scrubs the deleted row when the fetched page is still served from the stale cache. The service detail refetch is dropped: the only service-level field a delete can change is the cluster set, which the detail endpoint reads from the same lagging cache, so an immediate refetch cannot observe that change either.

Assisted-by: Claude Code

* [ISSUE #15296] Update console next static assets

Rebuilt console-ui-next static assets (npm ci && npm run build) for the instance toggle and delete fixes. No visual changes are included.

Assisted-by: Claude Code
(cherry picked from commit ab1ee0a)

#15336)

* feat:Update the disk cache by refreshing asynchronously

* fix:spotless:check

* fix: Add more tests to improve coverage

* fix: rollback changes to EnvUtil

* fix: Rebase develop

(cherry picked from commit e82c834)

(cherry picked from commit d54689a)

(cherry picked from commit 486a8d1)

* Fix MCP endpoint URL query preservation

* rebuild console-ui-next static assets

(cherry picked from commit 97d830b)

Assisted-by: Claude Code
(cherry picked from commit 4c79e2c)

* Refine skill diff and resource editing

* Add prompt version diff

* Sync next console static assets

# Conflicts:
#	console/src/main/resources/static/next/css/main3.css
#	console/src/main/resources/static/next/js/FileTreePanel.js
#	console/src/main/resources/static/next/js/index28.js
#	console/src/main/resources/static/next/js/index29.js
#	console/src/main/resources/static/next/js/index32.js
#	console/src/main/resources/static/next/js/index33.js
#	console/src/main/resources/static/next/js/index34.js
#	console/src/main/resources/static/next/js/index35.js
#	console/src/main/resources/static/next/js/main.js
#	console/src/main/resources/static/next/js/skill-store.js
#	console/src/main/resources/static/next/js/vendor-icons.js

* Handle non-text skill version diffs

* Refine AI resource diff interactions

* Fix AI resource file tree rename prop

* Sync next console static assets

* Sync next console static assets

(cherry picked from commit 5db14e3)

* Fix HTTP login TLS prefix

* style: apply spotless formatting

(cherry picked from commit a8d22aa)

* [ISSUE #15359] Fix AI registry version deletion

* style: apply spotless formatting

(cherry picked from commit ee86970)

* [ISSUE #12867] Add raft group metrics

* fix: preserve raft leader metrics on member events

(cherry picked from commit c42ec16)

* [ISSUE #12585] Fix pagination parameter binding

* style: apply spotless formatting

* test: align gray dump pagination parameters

(cherry picked from commit b886636)

…spec (#15445)

* fix: normalize null MCP endpoint protocol in console management flow

* fix: console-ui-new fix

(cherry picked from commit acd679c)

Move the sidebar collapse control to the header, place setting center under platform management, and add a user icon to the account dropdown.

Assisted-by: Claude Code
(cherry picked from commit e9e5aaa)

Reject blank instance IP through the Java SDK naming validation path and the server-side gRPC instance registration handler.

Add unit coverage, update Java SDK IT coverage for the blank-IP contract, and fix existing tests that intentionally need valid instance identities.

Assisted-by: Claude Code
(cherry picked from commit 7ec791b)

…15464)

* fix(ai): support nested skill resources in upload

* build: exclude generated and test-only files from checks

* build(console): update next static assets