v0.2.1

docker/README.md

@@ -112,7 +112,7 @@ The bundled SILO topology is intended for local/integration validation of the S3
 
 ### `compose-mysql-nfs-opensearch.yml`
 
-MySQL, shared POSIX/NFS blob storage and OpenSearch. The NFS volume driver is deliberate: a normal local Docker named volume is not shared between Docker hosts and therefore is not a valid blob backend for a multi-host RustBox deployment.
+MySQL, shared POSIX/NFS blob storage and OpenSearch. The NFS volume driver is deliberate: a normal local Docker named volume is not shared between Docker hosts and therefore is not a valid blob backend for a multi-host RustBox deployment. These examples intentionally run one RustBox application container; the shared backing services demonstrate storage/database topology, not multi-app load balancing.
 
 ## Production notes
 
@@ -135,3 +135,33 @@ ENVIRONMENT=production docker compose -p rustbox -f compose-postgres-s3-opensear
 In `production`, RustBox intentionally marks authentication cookies as `Secure`. Opening the application directly over plain HTTP in that mode causes the browser not to send the session cookie and protected requests return `Invalid or expired session`.
 
 After changing from production/HTTP to development/HTTP, log in again. If the browser still has stale RustBox state, clear cookies/site data for the RustBox origin once and reload.
+
+## OpenSearch disk watermark / 429 errors
+
+The `opensearch-data` named volume still consumes space on the filesystem used by Docker. Running OpenSearch in a container does not give it an independent disk. If the Docker host filesystem reaches OpenSearch's flood-stage threshold, OpenSearch protects indices with `read_only_allow_delete` and RustBox indexing receives HTTP 429 responses.
+
+Check the actual filesystem seen by the container:
+
+```bash
+docker compose -p rustbox -f compose-postgres-s3-opensearch.yml exec opensearch \
+  df -h /usr/share/opensearch/data
+```
+
+Also inspect Docker usage on the host:
+
+```bash
+df -h
+docker system df
+docker volume inspect rustbox_opensearch-data
+```
+
+Free disk space first. Do not permanently lower the OpenSearch flood-stage safety threshold. If the index remains read-only after disk usage is back below the high watermark, clear the block explicitly:
+
+```bash
+docker compose -p rustbox -f compose-postgres-s3-opensearch.yml exec opensearch \
+  curl -fsS -X PUT http://localhost:9200/rustbox-messages-v1/_settings \
+  -H 'Content-Type: application/json' \
+  -d '{"index.blocks.read_only_allow_delete":null}'
+```
+
+RustBox 0.2.1 backs off its OpenSearch outbox worker after these 429/flood-stage responses, so a full disk no longer produces one warning per queued event.

docker/compose-postgres-s3-opensearch.yml

@@ -76,7 +76,7 @@ services:
       SMTP_HOST: ${SMTP_HOST:-mail.example.com}
       SMTP_PORT: ${SMTP_PORT:-587}
       TLS_ENABLED: ${TLS_ENABLED:-true}
-      RUST_LOG: ${RUST_LOG:-info}
+      RUST_LOG: ${RUST_LOG:-info,sqlx::postgres::notice=warn}
       DATABASE_URL: postgres://rustbox:${POSTGRES_PASSWORD:-rustbox-dev-password}@postgres:5432/rustbox
       SESSION_ENCRYPTION_KEY: ${SESSION_ENCRYPTION_KEY:-MDEyMzQ1Njc4OWFiY2RlZjAxMjM0NTY3ODlhYmNkZWY=}
       AWS_ACCESS_KEY_ID: ${S3_ACCESS_KEY:-rustbox}