Repository navigation
fix: backport pre-release fixes to v1.0.x - #3586
Merged
Merged
Conversation
Upgrade tests on `main` now select `v1.0.0-rc3`, which already installs
Atelet as a separate Helm release. The regular upgrade test was blocked
by a missing optional migration table, then by missing Harness fixtures
when it tried to invoke the previous release.
Check for the optional tracking table in a separate statement:
PostgreSQL resolves table references before evaluating `CASE`. Keep
readiness polling for the required core migration table.
Before each invocation phase, apply the four Harness fixtures used by
`TestSessionInteraction` from the matching source checkout. Resolve
their images from the installed controller's runtime catalog so the
previous release, current build, and application rollback each exercise
their own runtime images. Reuse the existing lifecycle manifest without
creating its unrelated smoke Agents or opaque BYO fixture.
Validated with PostgreSQL query checks, fixture rendering tests for
released and locally built images, upgrade-package configuration tests,
and Go lint. Both required upgrade checks passed on `9bd2ab0fe`,
including invocation before and after the upgrade, schema comparison,
application rollback, and schema rollback.
<details>
<summary>Runtime fixtures across upgrade and rollback</summary>
```mermaid
flowchart LR
source["Matching release or PR checkout"] -->|Lifecycle manifest| fixtures["Invocation Harness fixtures"]
controller["Installed controller configuration"] -->|Runtime image digests| fixtures
fixtures -->|Apply| cluster["Upgrade test cluster"]
cluster -->|Invoke all four Harnesses| test["TestSessionInteraction"]
```
</details>
---------
Signed-off-by: Eitan Yarmush <eitan.yarmush@solo.io>
(cherry picked from commit 2e2acf5)
Signed-off-by: Eitan Yarmush <eitan.yarmush@solo.io>
…3545) --- *Written by 🤖 (start)* The Substrate page's actor and worker lists now show `Read at 4:41:02 PM` instead of `read 12.3s ago`, which re-rendered twice a second and pulled the eye off the rows. While polling, a list that goes 3 intervals (at least 5s) without a new read adds its age in the warning colour, e.g. `Read at 4:41:02 PM, 12s ago`. # Testing 1. Open Substrate. Each list shows `Read at <time>`, which changes only when the list refreshes. 2. Turn polling on, then make the API hang. After about 5s, each list reads `Read at <time>, <n>s ago` in amber. --- *Written by 🤖 (end)* --------- Signed-off-by: Nicholas Bucher <behappy54321@gmail.com> (cherry picked from commit 7c9c521) Signed-off-by: Eitan Yarmush <eitan.yarmush@solo.io>
## Summary When Substrate reports a Session's Actor `CRASHED`, a caller can revert it to its last durable snapshot through `SessionService.RevertSession`. Recovery runs only when a client asks: an admitted recovery is a durable `RUNTIME_OPERATION_REVERT` operation, and a retried `RevertSession` joins it, the same way Create and Delete retries work. ### How a revert works 1. **Observe.** `observeCrash` reads the Session's generation (lifecycle operation ID, recorded Actor UID, Actor template) in one statement, then reads the Actor from Substrate. A missing, replaced, or retemplated Actor is unrecoverable, never a crash. 2. **Admit.** `BeginCrashedSessionRecovery` runs one transaction under the Session lock. It: - checks the Actor UID and generation, and that no checkpoint is being created; - fails the interrupted turn with a timeline-positioned crash note, keeping waits the snapshot captured. Publishing saved results and retiring idle-work markers reuse the helpers Session deletion uses; - starts a REVERT operation. 3. **Execute.** The shared `execute` path reverts the Actor in place. Substrate restores the last completed durable snapshot. Every suspend of a kagent Actor is synchronous, so that is the snapshot admission settled against. `applyRevert` requires the same Actor back, SUSPENDED. If preparation fails, the REVERT is kept for a retry. ### API - `RevertSession` validates `session_id` as a UUID and needs `AccessUpdate`, so a read-only share cannot invoke it. - A runtime that is not crashed returns the Session unchanged. That is also what a retry sees after a revert whose response was lost. - A missing or replaced Actor returns `FAILED_PRECONDITION`. Contention returns `ABORTED`, and transient failures return `UNAVAILABLE`. - Delete supersedes a recovery even after a revert was issued. A revert restores the recorded Actor in place and creates nothing that Delete could miss. ### Schema `000002_session_revert_operation.sql` adds `RUNTIME_OPERATION_REVERT` to the `runtime_instance` operation CHECK. ## Testing - `make -C go lint`, `make proto-generate` - `go test` for `service/session`, `substrate`, `grpcserver` (revert), and `pkg/migrations` (up/down round trip) - PostgreSQL store tests, including `TestInlineSQLPrepares` - E2E: `go/core/test/e2e/session_revert_test.go` Stack: this PR → #3399 (GetSession recovery report and chat). 🤖 Generated with [Claude Code](https://claude.com/claude-code) --------- Signed-off-by: Ada Lundhe <adalundhe@lundhe.audio> Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> (cherry picked from commit aaf1ca8) Signed-off-by: Eitan Yarmush <eitan.yarmush@solo.io>
# Description Removing the KMCP integration and the `kagent mcp` scaffolder in #3223 left the shared CLI helpers it used without callers. This removes the Docker executor, project generator, filesystem, Kubernetes config, and input prompt helpers, trims the kubectl executor to what `kagent bug-report` still uses, and drops the `logging.FromLogr` adapter whose only caller was the MCPServer catalog check. No behavior change. The remaining `deadcode` findings are public STS helpers and the `WithGRPCTransportCredentials` client option, which are intentionally kept as consumer-facing extension points. Signed-off-by: timflannagan <timflannagan@gmail.com> (cherry picked from commit 5c490d8) Signed-off-by: Eitan Yarmush <eitan.yarmush@solo.io>
# Description `go/api/v1alpha3` is the public API that external clients and wrapper types import, but it pulled in controller-runtime and `database/sql` for a few helpers that do I/O. This moves those helpers out so the API package only describes data. `RemoteMCPServerSpec` loses its `sql.Scanner`/`driver.Valuer` methods, which nothing used. `RemoteMCPServer.ResolveHeaders`, `ValueRef.Resolve`, and `ValueSource.Resolve` move into the tool service's MCP client, their only caller, and `go/api/utils` is removed. The translator keeps its existing `HeadersFrom` handling, which compiles Secret-backed headers into egress credentials rather than reading them. Header resolution behaves the same, except errors are now wrapped and name the failing header. Go consumers that called the removed methods need to read the referenced Secret or ConfigMap themselves. Signed-off-by: timflannagan <timflannagan@gmail.com> (cherry picked from commit c6adf72) Signed-off-by: Eitan Yarmush <eitan.yarmush@solo.io>
…on a status change (#3340) ## Problem The ModelConfig status writer moves a condition's `lastTransitionTime` whenever its reason, message or observedGeneration changes. Agent conditions, and SandboxTemplate through `apimeta.SetStatusCondition`, move it only when the status changes, as the API conventions define it. A ModelConfig that stays `False` therefore reports a new transition time every time its error text or generation changes. ## Change Agent and ModelConfig share one helper that keeps the transition time while a condition's status holds. A new test covers a changed reason, message and generation under an unchanged status, and a status change. This PR previously wrote Harness `Ready`/`Booted` conditions; #3454 removed `Harness.status`, so it is reduced to the transition-time rule that part shared with Agent and ModelConfig. Signed-off-by: QuentinBisson <quentin@giantswarm.io> (cherry picked from commit 347acbe) Signed-off-by: Eitan Yarmush <eitan.yarmush@solo.io>
# Description Slow MCP discovery can exceed router timeouts and fail agent resumes, holding worker reservations and capacity while startup waits. Separate classification also duplicates MCP connections and paginated listings. This fixes it by: - Deferring HTTP, SSE, and stdio MCP connections until invocation preparation. - Classifying and filtering ADK's own discovery responses before the SDK caches them, preserving server TTLs and catalog notification handling. - Omitting failed toolsets for the execution regardless of prior discovery success, keeping healthy tools available and retrying on later executions without exposing unclassified tools. - Serializing discovery and classification publication per toolset, with cancellation support for queued invocations. - Preserving MCP App result compaction through a single callback, alongside rich stored results, credentials, approvals, and tracing. Tools reclassified as ordinary keep their real output; removed app names remain available for historical compaction. - Documenting discovery, SDK caching, pagination guards, and failure handling in `docs/architecture/mcp.md`. # Change Type ``` /kind fix ``` # Testing - Added unit coverage for zero startup requests, one paginated discovery, filtered SDK cache hits, TTL expiry, catalog notifications, partial failures, execution-scoped retries after initial and later failures, cancellation, concurrent discovery, reclassification, and stdio startup. - Added app-result coverage for cached tools, approvals, and historical results after catalog removal, preserving the original rich payloads. - Added E2E coverage for slow MCP discovery across repeated resumes and initial discovery failure and recovery; extended app-history E2E coverage to verify SDK caching across model steps and rediscovery after restore. # Changelog ```release-note MCP discovery no longer delays startup or resume and caches filtered tools safely. ``` --------- Signed-off-by: Shashank Ram <shashank.ram@solo.io> (cherry picked from commit ccb2dd3) Signed-off-by: Eitan Yarmush <eitan.yarmush@solo.io>
EItanya
disabled auto-merge
November 9, 2026 14:18
github-actions
Bot
added
bug
EItanya
enabled auto-merge
November 9, 2026 14:34
This comment has been minimized.
This comment has been minimized.
jjamroga
approved these changes
Nov 9, 2026
timflannagan
approved these changes
Nov 9, 2026
github-merge-queue
Bot
removed this pull request from the merge queue due to failed status checks
Nov 9, 2026
This was referenced Nov 9, 2026
This was referenced Nov 9, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Backports seven PRs from
maintorelease/v1.0.xfor the 1.0 release. Each is cherry-picked with-xin its original merge order, and none needed changes. Every file these commits touch now matchesmainexactly.2e2acf58c7c9c52105aaf1ca8825c490d83bmain)c6adf72f9v1alpha3, so the 1.0 API package describes only data347acbe55cherrpick-1.0)ccb2dd350Please use Rebase and merge so each backported PR stays its own commit on the release branch.
Validation before #3560 was added (CI covers the full set):
go build ./...andgo vet ./...go test -racewith envtest forapi,core/cli,core/internal/{controller,database,grpcserver,mcp,service,substrate},core/pkg/migrations, andpkg/loggingmake -C go lint: 0 issuestsc --noEmitandsubstrateReadAge.test.tsx🤖 Generated with Claude Code