@blinklabs/dingo 0.27.2 → 0.27.4

package/RELEASE_NOTES.md

@@ -1,6 +1,90 @@
 # Release Notes
 
 
+## v0.27.3 (March 17, 2026)
+
+**Title:** Safer rollbacks and steadier leader election
+
+**Date:** March 17, 2026
+
+**Version:** v0.27.3
+
+Hi folks! Here’s what we shipped in v0.27.3.
+
+### ✨ What's New
+
+- **Follow-tip reset and rollback:** Tip tracking is easier to manage because the follow-tip API now supports safe reset and rollback with clearer metadata about what changed.
+- **Leader election readiness:** Block producer readiness is easier to track because leader election now surfaces epoch-nonce readiness and carries schedule state more reliably across restarts.
+
+### 💪 Improvements
+
+- **More consistent cache sizing:** Sizing runs are easier to tune because cache defaults are more consistent and the BP/PI sizing script now supports explicit memory limits and optional cache overrides.
+- **Predictable KES period semantics:** Operational certificate (KES) periods are more predictable because KES endpoints now standardize on absolute periods while translating internally from the certificate start period.
+- **Preserved original block bytes:** Downstream tooling can retain exact block bytes because API block objects now include the original encoded bytes.
+- **Standard network identifiers from genesis:** Network identification is simpler because genesis reads now return a standard CAIP-2 network identifier derived from network magic.
+- **Smoother peer switching:** Sync stays more rock-solid because chain-sync now preserves its state while only swapping the active connection during a peer switch.
+- **More consistent Mithril imports:** Mithril snapshot imports are more consistent across epochs because imports now normalize snapshot types and centralize persistence and epoch-summary handling.
+- **Protocol dependency validation:** Modern-era transaction handling is more reliable because protocol dependencies were updated and regression tests now guard transaction size behavior.
+- **Safer default containers:** Default containers are safer because the main Docker image now runs as a non-root `dingo` user.
+
+### 🔧 Fixes
+
+- **Resilient background monitoring:** Long-running monitoring is more rock-solid because the stall checker now recovers from panics instead of crashing its background loop.
+- **Robust block fetch batching:** Block fetch serving is more robust because batching now handles iterator errors and connection closes correctly.
+- **Clearer unexpected event handling:** Event processing is easier to debug because chain-sync and block fetch now log unexpected event payload types instead of failing silently.
+- **Reliable shutdown error reporting:** Shutdowns are easier to troubleshoot because node and metrics server shutdown now propagates errors instead of exiting abruptly.
+- **Clear tx-submission failures:** Transaction submission fails more clearly because the tx-submission handlers now return an explicit error when no mempool consumer exists.
+
+### 📋 What You Need to Know
+
+- **Docker volume permissions:** If you run Dingo in Docker with mounted volumes, make sure the data directory is writable by the `dingo` user inside the container.
+- **API integrations:** If you integrate with follow-tip or KES APIs, give your client code a quick check for the updated reset/rollback and period semantics.
+
+### 🙏 Thank You
+
+Thank you for trying!
+
+---
+
+## v0.27.2 (March 16, 2026)
+
+**Title:** Snapshot events and safer services
+
+**Date:** March 16, 2026
+
+**Version:** v0.27.2
+
+Hi folks! Here’s what we shipped in v0.27.2.
+
+### ✨ What's New
+
+- **Snapshot event publishing and clean shutdown:** Relay operation is more reliable because the relay now publishes events from snapshots and shuts down cleanly without dropping in-flight work.
+
+### 💪 Improvements
+
+- **HTTP timeouts for public APIs:** Network-facing services are more resilient under slow or stalled connections thanks to new write/read/idle timeouts on the Bark, Blockfrost, and UTxO RPC HTTP servers.
+- **Streamlined peer and connection management:** Peer and connection management is faster and uses fewer resources on constrained machines thanks to quicker inbound host lookups, tighter Badger cache defaults, and expanded benchmarks and sizing guidance.
+- **More consistent key-period handling:** Block production key period handling is more consistent across configurations thanks to improved key-period calculations with added validation and tests.
+- **Race detection in CI:** Test runs catch concurrency bugs earlier because the Linux test job now runs with the Go race detector enabled.
+
+### 🔧 Fixes
+
+- **Safer concurrent chain reads:** Reads are more consistent under load because primary chain and protocol-parameter access are now protected with read locks.
+- **Validated epoch nonce reuse:** Nonce reuse is safer because cached epoch nonce entries are now validated against the nonce provided for the current run before reuse.
+- **Graceful invalid hash handling:** Malformed block metadata no longer crashes encoding because previous-hash length issues now return errors instead of panicking.
+- **SQLite VACUUM actually runs:** Database maintenance now completes as intended because SQLite VACUUM is now executed rather than only prepared.
+
+### 📋 What You Need to Know
+
+- **Release notes alignment:** Release documentation was updated to reflect the final set of changes, including a detailed v0.27.1 section in `RELEASE_NOTES.md`.
+- **Build provenance updates:** Supply-chain attestations are easier to verify because build provenance now uses `actions/attest` and updated Docker Hub image subjects.
+
+### 🙏 Thank You
+
+Thank you for trying!
+
+---
+
 ## v0.27.1 (March 16, 2026)
 
 **Title:** Smoother reconnects and safer chain-sync

package/generate_bp_pi_report.sh

@@ -8,6 +8,11 @@ CACHE_DIR="${CACHE_DIR:-/tmp/dingo-bp-pi-go-cache}"
 GOMAXPROCS_VALUE="${GOMAXPROCS_VALUE:-4}"
 BENCH_TIME="${BENCH_TIME:-5s}"
 IMMUTABLE_PATH="${IMMUTABLE_PATH:-database/immutable/testdata}"
+RSS_LIMIT_MB="${RSS_LIMIT_MB:-1024}"
+GOMEMLIMIT_VALUE="${GOMEMLIMIT_VALUE:-}"
+BLOCK_CACHE_SIZE_VALUE="${BLOCK_CACHE_SIZE_VALUE:-}"
+INDEX_CACHE_SIZE_VALUE="${INDEX_CACHE_SIZE_VALUE:-}"
+MEMTABLE_SIZE_VALUE="${MEMTABLE_SIZE_VALUE:-}"
 DEFAULT_DB_DIR="${DEFAULT_DB_DIR:-}"
 DEFAULT_LOG_FILE="${DEFAULT_LOG_FILE:-}"
 DB_DIR_BASE="${DB_DIR_BASE:-/tmp}"
@@ -16,6 +21,7 @@ TIME_FORMAT_STYLE=""
 TIME_CMD=()
 AUTO_DEFAULT_DB_DIR=0
 AUTO_DEFAULT_LOG_FILE=0
+DINGO_CACHE_ARGS=()
 
 cleanup() {
   local exit_status=$?
@@ -197,18 +203,25 @@ run_load_profile() {
   validate_db_dir "$db_dir" "$log_file"
   rm -rf "$db_dir"
   : > "$log_file"
+  local -a env_args=(
+    PATH="$PATH"
+    HOME="${HOME:-/tmp}"
+    GOMAXPROCS="$GOMAXPROCS_VALUE"
+    DINGO_RUN_MODE=serve
+    DINGO_STORAGE_MODE=core
+  )
+  if [[ -n "$GOMEMLIMIT_VALUE" ]]; then
+    env_args+=(GOMEMLIMIT="$GOMEMLIMIT_VALUE")
+  fi
   env -i \
-    PATH="$PATH" \
-    HOME="${HOME:-/tmp}" \
-    GOMAXPROCS="$GOMAXPROCS_VALUE" \
-    DINGO_RUN_MODE=serve \
-    DINGO_STORAGE_MODE=core \
+    "${env_args[@]}" \
     "${TIME_CMD[@]}" \
     ./dingo \
     --config /dev/null \
     --blob badger \
     --metadata sqlite \
     --data-dir "$db_dir" \
+    "${DINGO_CACHE_ARGS[@]}" \
     load "$IMMUTABLE_PATH" "$@" \
     > "$log_file" 2>&1
 
@@ -231,6 +244,16 @@ parse_metric() {
 
 configure_time_cmd
 
+if [[ -n "$BLOCK_CACHE_SIZE_VALUE" ]]; then
+  DINGO_CACHE_ARGS+=(--block-cache-size "$BLOCK_CACHE_SIZE_VALUE")
+fi
+if [[ -n "$INDEX_CACHE_SIZE_VALUE" ]]; then
+  DINGO_CACHE_ARGS+=(--index-cache-size "$INDEX_CACHE_SIZE_VALUE")
+fi
+if [[ -n "$MEMTABLE_SIZE_VALUE" ]]; then
+  DINGO_CACHE_ARGS+=(--memtable-size "$MEMTABLE_SIZE_VALUE")
+fi
+
 DB_DIR_BASE="$(resolve_path "$DB_DIR_BASE")"
 
 if [[ -z "$DEFAULT_DB_DIR" ]]; then
@@ -279,6 +302,12 @@ if [[ ! "$default_rss_kb" =~ ^[0-9]+$ ]]; then
 fi
 
 default_rss_mb="$(awk "BEGIN { printf \"%.1f\", ${default_rss_kb} / 1024 }")"
+rss_limit_kb=$((RSS_LIMIT_MB * 1024))
+
+if (( default_rss_kb > rss_limit_kb )); then
+  echo "error: peak RSS ${default_rss_mb} MB exceeds limit ${RSS_LIMIT_MB} MB" >&2
+  exit 1
+fi
 
 {
   echo "# Four-Thread Block Producer Sizing Report (Non-Pi Host)"
@@ -287,7 +316,23 @@ default_rss_mb="$(awk "BEGIN { printf \"%.1f\", ${default_rss_kb} / 1024 }")"
   printf -- "- \`storageMode=core\`\n"
   printf -- "- \`runMode=serve\`, \`blob=badger\`, \`metadata=sqlite\`\n"
   printf -- "- \`GOMAXPROCS=%s\`\n" "$GOMAXPROCS_VALUE"
+  printf -- "- target peak RSS: \`<= %s MB\`\n" "$RSS_LIMIT_MB"
   printf -- "- immutable fixture: \`%s\`\n" "$IMMUTABLE_PATH"
+  if [[ -n "$GOMEMLIMIT_VALUE" ]]; then
+    printf -- "- test-only \`GOMEMLIMIT=%s\`\n" "$GOMEMLIMIT_VALUE"
+  fi
+  if [[ -n "$BLOCK_CACHE_SIZE_VALUE" || -n "$INDEX_CACHE_SIZE_VALUE" || -n "$MEMTABLE_SIZE_VALUE" ]]; then
+    echo "- explicit Badger test-only overrides:"
+    if [[ -n "$BLOCK_CACHE_SIZE_VALUE" ]]; then
+      printf -- "  - \`--block-cache-size=%s\`\n" "$BLOCK_CACHE_SIZE_VALUE"
+    fi
+    if [[ -n "$INDEX_CACHE_SIZE_VALUE" ]]; then
+      printf -- "  - \`--index-cache-size=%s\`\n" "$INDEX_CACHE_SIZE_VALUE"
+    fi
+    if [[ -n "$MEMTABLE_SIZE_VALUE" ]]; then
+      printf -- "  - \`--memtable-size=%s\`\n" "$MEMTABLE_SIZE_VALUE"
+    fi
+  fi
   echo "- measurements were collected on non-Pi hardware and are a sizing proxy, not Raspberry Pi device benchmarks"
   echo "- block producer peer count remains small; this is not a relay profile"
   echo
@@ -295,11 +340,16 @@ default_rss_mb="$(awk "BEGIN { printf \"%.1f\", ${default_rss_kb} / 1024 }")"
   echo
   echo "| Profile | Elapsed | Peak RSS | Notes |"
   echo "| --- | --- | --- | --- |"
-  printf -- "| \`core-default\` | \`%ss\` | \`%s MB\` | Current default \`serve/core\` cache profile |\n" "$default_elapsed" "$default_rss_mb"
+  if [[ -n "$BLOCK_CACHE_SIZE_VALUE" || -n "$INDEX_CACHE_SIZE_VALUE" || -n "$MEMTABLE_SIZE_VALUE" || -n "$GOMEMLIMIT_VALUE" ]]; then
+    printf -- "| \`core-constrained\` | \`%ss\` | \`%s MB\` | Explicit test-only memory/cache overrides |\n" "$default_elapsed" "$default_rss_mb"
+  else
+    printf -- "| \`core-default\` | \`%ss\` | \`%s MB\` | Current default \`serve/core\` cache profile |\n" "$default_elapsed" "$default_rss_mb"
+  fi
   echo
   echo "## Recommendation"
   echo
-  printf -- "For Raspberry Pi-class sizing in \`storageMode=core\`, the current default \`serve/core\` profile is already in the measured low-memory range on this four-thread non-Pi host.\n"
+  printf -- "For Raspberry Pi-class sizing in \`storageMode=core\`, keep production defaults aligned with normal Badger behavior and use explicit memory/cache constraints only in this harness when validating low-memory targets.\n"
+  printf -- "This run stayed under the configured \`%s MB\` RSS limit.\n" "$RSS_LIMIT_MB"
   echo "Explicit Badger cache settings still override these defaults if an operator wants to tune further."
   echo "Treat this as a sizing baseline rather than a direct Raspberry Pi hardware measurement."
   echo

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@blinklabs/dingo",
-  "version": "0.27.2",
+  "version": "0.27.4",
   "description": "Dingo is a Cardano blockchain data node",
   "main": "index.js",
   "bin": {