From f44acd0974d9bcf65eb1b5b37023fa0b17be74c3 Mon Sep 17 00:00:00 2001
From: Tom Gundersen <teg@jklm.no>
Date: Mon, 25 Oct 2021 16:45:45 +0100
Subject: [PATCH] Revert "Revert "cloudapi/v1: Move depsolving to workers""

Workers now depsolve in parallel to image builds, so we can
again move depsolivng to the workers. This will help us deal
with increases in traffic as we currently only have one
depsolve handler per pod. It would also avoid any issues with
composer running out of disk space due to dnf metadata caches.

This reverts commit c65b1e9b26779fd65f724d3abe7288b7d366a265.
---
 internal/cloudapi/v1/v1.go | 50 +++++++++++++++++++++++++-------------
 test/cases/api.sh          | 12 ++-------
 2 files changed, 35 insertions(+), 27 deletions(-)

diff --git a/internal/cloudapi/v1/v1.go b/internal/cloudapi/v1/v1.go
index 1786e1481..920034bd1 100644
--- a/internal/cloudapi/v1/v1.go
+++ b/internal/cloudapi/v1/v1.go
@@ -10,6 +10,7 @@ import (
 	"math/big"
 	"net/http"
 	"strings"
+	"time"
 
 	"github.com/google/uuid"
 	"github.com/labstack/echo/v4"
@@ -157,25 +158,40 @@ func (h *apiHandlers) Compose(ctx echo.Context) error {
 		}
 
 		packageSets := imageType.PackageSets(bp)
-		pkgSpecSets := make(map[string][]rpmmd.PackageSpec)
-		for name, packages := range packageSets {
-			pkgs, _, err := h.server.rpmMetadata.Depsolve(packages, repositories, distribution.ModulePlatformID(), arch.Name(), distribution.Releasever())
-			if err != nil {
-				var error_type int
-				switch err.(type) {
-				// Known DNF errors falls under BadRequest
-				case *rpmmd.DNFError:
-					error_type = http.StatusBadRequest
-				// All other kind of errors are internal server Errors.
-				// (json marshalling issues for instance)
-				case error:
-					error_type = http.StatusInternalServerError
-				}
-				return echo.NewHTTPError(error_type, "Failed to depsolve base packages for %s/%s/%s: %s", ir.ImageType, ir.Architecture, request.Distribution, err)
-			}
-			pkgSpecSets[name] = pkgs
+		depsolveJobID, err := h.server.workers.EnqueueDepsolve(&worker.DepsolveJob{
+			PackageSets:      packageSets,
+			Repos:            repositories,
+			ModulePlatformID: distribution.ModulePlatformID(),
+			Arch:             arch.Name(),
+			Releasever:       distribution.Releasever(),
+		})
+		if err != nil {
+			return echo.NewHTTPError(http.StatusInternalServerError, "Unable to enqueue depsolve job")
 		}
 
+		var depsolveResults worker.DepsolveJobResult
+		for {
+			status, _, err := h.server.workers.JobStatus(depsolveJobID, &depsolveResults)
+			if err != nil {
+				return echo.NewHTTPError(http.StatusInternalServerError, "Unable to get depsolve results")
+			}
+			if status.Canceled {
+				return echo.NewHTTPError(http.StatusInternalServerError, "Depsolving job canceled unexpectedly")
+			}
+			if !status.Finished.IsZero() {
+				break
+			}
+			time.Sleep(50 * time.Millisecond)
+		}
+
+		if depsolveResults.Error != "" {
+			if depsolveResults.ErrorType == worker.DepsolveErrorType {
+				return echo.NewHTTPError(http.StatusBadRequest, "Failed to depsolve requested package set: %s", depsolveResults.Error)
+			}
+			return echo.NewHTTPError(http.StatusInternalServerError, "Error while depsolving: %s", depsolveResults.Error)
+		}
+		pkgSpecSets := depsolveResults.PackageSpecs
+
 		imageOptions := distro.ImageOptions{Size: imageType.Size(0)}
 		if request.Customizations != nil && request.Customizations.Subscription != nil {
 			imageOptions.Subscription = &distro.SubscriptionImageOptions{
diff --git a/test/cases/api.sh b/test/cases/api.sh
index 731aa761f..8daa388c3 100755
--- a/test/cases/api.sh
+++ b/test/cases/api.sh
@@ -706,17 +706,9 @@ function waitForState() {
     done
 }
 
-# a pending shouldn't state shouldn't trip up the heartbeats
-sudo systemctl stop "osbuild-worker@*"
 sendCompose
-waitForState "pending"
-# jobs time out after 2 minutes, so 180 seconds gives ample time to make sure it
-# doesn't time out for pending jobs
-sleep 180
-waitForState "pending"
 
 # crashed/stopped/killed worker should result in a failed state
-sudo systemctl start "osbuild-worker@1"
 waitForState "building"
 sudo systemctl stop "osbuild-worker@*"
 waitForState "failure"
@@ -734,8 +726,8 @@ test "$UPLOAD_STATUS" = "success"
 test "$UPLOAD_TYPE" = "$CLOUD_PROVIDER"
 test $((INIT_COMPOSES+1)) = "$SUBS_COMPOSES"
 
-# Make sure we get 1 job entry in the db per compose
-sudo podman exec osbuild-composer-db psql -U postgres -d osbuildcomposer -c "SELECT * FROM jobs;" | grep "2 rows"
+# Make sure we get 2 job entries in the db per compose (depsolve + build)
+sudo podman exec osbuild-composer-db psql -U postgres -d osbuildcomposer -c "SELECT * FROM jobs;" | grep "4 rows"
 
 #
 # Save the Manifest from the osbuild-composer store