Merge branch 'main' of github.com:e2b-dev/infra into lev-block-cache-bitmap

Author: levb

.github/actions/build-sandbox-template/action.yml

@@ -8,7 +8,7 @@ runs:
       env:
         TEMPLATE_ID: "2j6ly824owf4awgai1xo"
         KERNEL_VERSION: "vmlinux-6.1.158"
-        FIRECRACKER_VERSION: "v1.12.1_a41d3fb"
+        FIRECRACKER_VERSION: "v1.12.1_210cbac"
       run: |
         # Generate an unique build ID for the template for this run
         export BUILD_ID=$(uuidgen)

.github/workflows/periodic-test.yml

@@ -103,56 +103,62 @@ jobs:
           E2B_ACCESS_TOKEN: ${{ env.E2B_ACCESS_TOKEN }}
           E2B_DOMAIN: ${{ matrix.e2b-domain }}
 
-      - name: Send Grafana OnCall webhook
+      - name: Send incident.io alert
         if: always()
+        continue-on-error: true
         env:
           JOB_STATUS: ${{ job.status }}
           DOMAIN: ${{ matrix.e2b-domain }}
           RUN_LINK: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}
-          WEBHOOK_URL: ${{ secrets.GRAFANA_ALERT_WEBHOOK }}
+          WEBHOOK_URL: ${{ secrets.INCIDENT_IO_ALERT_SOURCE_URL }}
+          ALERT_TOKEN: ${{ secrets.INCIDENT_IO_ALERT_TOKEN }}
           TEST_NAME: ${{ matrix.test-command.name }}
         run: |
-          sudo apt-get update -qq && sudo apt-get install -y -qq jq  # jq for JSON assembly
+          sudo apt-get update -qq && sudo apt-get install -y -qq jq
 
-          # Map GH status → Grafana state
-          STATE=$([ "$JOB_STATUS" = "success" ] && echo ok || echo alerting)
+          # Map GH status to incident.io status
           STATUS=$([ "$JOB_STATUS" = "success" ] && echo resolved || echo firing)
 
-          # slug-ify the test name (lowercase, a–z/0–9, hyphens)
+          # slug-ify the test name (lowercase, a-z/0-9, hyphens)
           TEST_SLUG=$(echo "$TEST_NAME" \
                       | tr '[:upper:]' '[:lower:]' \
                       | sed -E 's/[^a-z0-9]+/-/g; s/^-+|-+$//g')
 
-          # Unique per (domain + test) – same UID for state so OnCall auto-resolves
-          ALERT_UID="periodic-tests-${DOMAIN//./-}-${TEST_SLUG}"
+          # Unique per (domain + test) – same key so incident.io auto-resolves
+          DEDUP_KEY="periodic-tests-${DOMAIN//./-}-${TEST_SLUG}"
 
           TITLE="Periodic tests | ${DOMAIN} | ${TEST_NAME}"
-          MESSAGE="${TEST_NAME} on ${DOMAIN} finished. Link: ${RUN_LINK}"
+          DESCRIPTION="${TEST_NAME} on ${DOMAIN} finished with status: ${JOB_STATUS}. [View run](${RUN_LINK})"
 
-          # Attach a branch-status badge.
+          # Attach a branch-status badge
           BADGE_COLOR=$([ "$JOB_STATUS" = "success" ] && echo brightgreen || echo CB2431)
-          BADGE_LABEL_ESCAPED=$(echo "$ALERT_UID" | sed 's/-/--/g')
+          BADGE_LABEL_ESCAPED=$(echo "$DEDUP_KEY" | sed 's/-/--/g')
           IMAGE_URL="https://img.shields.io/badge/${BADGE_LABEL_ESCAPED}-${JOB_STATUS}-${BADGE_COLOR}"
 
           jq -n \
-            --arg alert_uid "$ALERT_UID" \
+            --arg deduplication_key "$DEDUP_KEY" \
             --arg title "$TITLE" \
-            --arg state "$STATE" \
             --arg status "$STATUS" \
-            --arg link_to_upstream_details "$RUN_LINK" \
-            --arg message "$MESSAGE" \
-            --arg image_url "$IMAGE_URL" \
+            --arg source_url "$RUN_LINK" \
+            --arg description "$DESCRIPTION" \
             --arg domain "$DOMAIN" \
             --arg dev "${{ inputs.dev }}" \
-            '{alert_uid: $alert_uid,
+            --arg image_url "$IMAGE_URL" \
+            '{
+              deduplication_key: $deduplication_key,
               title: $title,
-              state: $state,
-              link_to_upstream_details: $link_to_upstream_details,
-              message: $message,
-              severity: "warning",
               status: $status,
-              domain: $domain,
-              dev: ($dev == "true"),
-              annotations: {Project: $domain},
-              image_url: $image_url}' \
-          | curl -s -X POST -H 'Content-Type: application/json' -d @- "$WEBHOOK_URL"
+              source_url: $source_url,
+              description: $description,
+              metadata: {
+                severity: "warning",
+                domain: $domain,
+                dev: ($dev == "true"),
+                image_url: $image_url,
+                project: $domain
+              }
+            }' \
+          | curl -s --fail-with-body -X POST \
+              -H 'Content-Type: application/json' \
+              -H "Authorization: Bearer $ALERT_TOKEN" \
+              -d @- "$WEBHOOK_URL"

iac/modules/job-otel-collector/configs/otel-collector.yaml

@@ -230,6 +230,8 @@ processors:
           - "pgxpool.*"
           - "go\\..*"
           - "traefik.*"
+          - "grpc\\..*"
+          - "rpc\\.client.*"
 
   filter/only_client_allocs:
     metrics:

iac/provider-gcp/Makefile

@@ -111,10 +111,6 @@ init:
 .PHONY: state-migrate
 state-migrate:
 	@ printf "Running Terraform state migrations for env: `tput setaf 2``tput bold`$(ENV)`tput sgr0`\n\n"
-	@ $(tf_vars) $(TF) state mv 'module.nomad.nomad_job.api' 'module.nomad.module.api.nomad_job.api' || true
-	@ $(tf_vars) $(TF) state mv 'module.nomad.nomad_job.template_manager' 'module.nomad.module.template_manager.nomad_job.template_manager' || true
-	@ $(tf_vars) $(TF) state mv 'module.nomad.nomad_job.nomad_nodepool_apm[0]' 'module.nomad.module.template_manager_autoscaler[0].nomad_job.nomad_nodepool_apm' || true
-	@ $(tf_vars) $(TF) state mv 'module.nomad.nomad_job.redis[0]' 'module.nomad.module.redis[0].nomad_job.redis' || true
 
 .PHONY: apply-init
 apply-init:

packages/api/go.mod

@@ -54,7 +54,7 @@ require (
 	go.opentelemetry.io/otel/trace v1.41.0
 	go.uber.org/zap v1.27.1
 	golang.org/x/net v0.50.0
-	golang.org/x/sync v0.19.0
+	golang.org/x/sync v0.20.0
 	google.golang.org/grpc v1.79.3
 	google.golang.org/protobuf v1.36.11
 )
@@ -380,11 +380,11 @@ require (
 	golang.org/x/arch v0.18.0 // indirect
 	golang.org/x/crypto v0.48.0 // indirect
 	golang.org/x/exp v0.0.0-20260212183809-81e46e3db34a // indirect
-	golang.org/x/image v0.25.0 // indirect
+	golang.org/x/image v0.38.0 // indirect
 	golang.org/x/mod v0.33.0 // indirect
 	golang.org/x/oauth2 v0.34.0 // indirect
 	golang.org/x/sys v0.41.0 // indirect
-	golang.org/x/text v0.34.0 // indirect
+	golang.org/x/text v0.35.0 // indirect
 	golang.org/x/time v0.14.0 // indirect
 	golang.org/x/tools v0.42.0 // indirect
 	google.golang.org/api v0.257.0 // indirect

packages/api/go.sum

@@ -6,25 +6,25 @@ import (
 	"crypto/x509"
 	"fmt"
 
+	"go.opentelemetry.io/contrib/instrumentation/google.golang.org/grpc/otelgrpc"
 	"google.golang.org/grpc"
 	"google.golang.org/grpc/credentials"
 	"google.golang.org/protobuf/types/known/emptypb"
 
-	"github.com/e2b-dev/infra/packages/shared/pkg/logger"
+	e2bgrpc "github.com/e2b-dev/infra/packages/shared/pkg/grpc"
 )
 
 type Analytics struct {
 	client     AnalyticsCollectorClient
 	connection *grpc.ClientConn
 }
 
-func NewAnalytics(ctx context.Context, host, grpcAPIKey string) (*Analytics, error) {
+func NewAnalytics(host, grpcAPIKey string) (*Analytics, error) {
 	var client AnalyticsCollectorClient
 	var connection *grpc.ClientConn
 
-	if host == "" {
-		logger.L().Info(ctx, "Running dummy implementation of analytics collector client, no host provided")
-	} else {
+	// Run dummy client if host is not provided
+	if host != "" {
 		systemRoots, err := x509.SystemCertPool()
 		if err != nil {
 			errMsg := fmt.Errorf("failed to read system root certificate pool: %w", err)
@@ -39,6 +39,7 @@ func NewAnalytics(ctx context.Context, host, grpcAPIKey string) (*Analytics, err
 
 		conn, err := grpc.NewClient(
 			fmt.Sprintf("%s:443", host),
+			grpc.WithStatsHandler(otelgrpc.NewClientHandler()),
 			grpc.WithPerRPCCredentials(newGRPCAPIKey(grpcAPIKey)),
 			grpc.WithAuthority(host),
 			grpc.WithTransportCredentials(cred),
@@ -67,6 +68,10 @@ func (a *Analytics) Close() error {
 	return nil
 }
 
+func (a *Analytics) Init(ctx context.Context) {
+	e2bgrpc.ObserveConnection(ctx, a.connection, "analytics-collector")
+}
+
 func (a *Analytics) InstanceStarted(ctx context.Context, in *InstanceStartedEvent, opts ...grpc.CallOption) (*emptypb.Empty, error) {
 	if a.client == nil {
 		return nil, nil