Files
opencloud-compose/search/opensearch.yml
Dominik Schmidt c814f08cd5 feat: add opensearch search backend module
Adds search/opensearch.yml — a single-node OpenSearch deployment wired up
as the opencloud search engine backend via SEARCH_ENGINE_TYPE=open-search.

The security plugin is disabled because the service is only reachable on
the internal opencloud-net bridge (no published port). Do not publish or
proxy this without enabling the security plugin first.

The disk-based shard allocation watermarks are disabled so the index
does not flip to read-only on modest disks, which otherwise silently
turns bulk upserts into no-ops.
2026-04-22 00:43:41 +02:00

60 lines
2.2 KiB
YAML

---
services:
opencloud:
environment:
# Point the search service at OpenSearch instead of the embedded bleve index.
SEARCH_ENGINE_TYPE: open-search
SEARCH_ENGINE_OPEN_SEARCH_CLIENT_ADDRESSES: http://opensearch:9200
SEARCH_ENGINE_OPEN_SEARCH_CLIENT_INSECURE: "true"
SEARCH_ENGINE_OPEN_SEARCH_RESOURCE_INDEX_NAME: ${OPENSEARCH_RESOURCE_INDEX:-opencloud-resources}
depends_on:
opensearch:
condition: service_healthy
opensearch:
image: ${OPENSEARCH_DOCKER_IMAGE:-opensearchproject/opensearch}:${OPENSEARCH_DOCKER_TAG:-2.19.5}
environment:
discovery.type: single-node
bootstrap.memory_lock: "true"
OPENSEARCH_JAVA_OPTS: ${OPENSEARCH_JAVA_OPTS:--Xms512m -Xmx512m}
# Security plugin is disabled: OpenSearch is only reachable on the
# internal opencloud-net bridge and no port is published to the host.
# Do NOT enable a published port or expose this via the reverse proxy
# without first enabling and configuring the security plugin.
DISABLE_SECURITY_PLUGIN: "true"
DISABLE_INSTALL_DEMO_CONFIG: "true"
# Disable the disk-based shard allocation watermarks. By default OpenSearch
# marks indices read-only when the host disk is <5% free, which silently
# turns bulk upserts into no-ops. Fine to disable on a single-node dev box.
cluster.routing.allocation.disk.threshold_enabled: "false"
ulimits:
memlock:
soft: -1
hard: -1
nofile:
soft: 65536
hard: 65536
networks:
opencloud-net:
volumes:
- ${OPENSEARCH_DATA_DIR:-opensearch-data}:/usr/share/opensearch/data
healthcheck:
# Single-node clusters can't go green (replicas have nowhere to land), so
# yellow is the healthy state. Still gates opencloud until opensearch accepts
# requests.
test:
[
"CMD-SHELL",
"curl -sf 'http://localhost:9200/_cluster/health?wait_for_status=yellow&timeout=5s' > /dev/null || exit 1",
]
interval: 5s
timeout: 10s
retries: 24
start_period: 60s
logging:
driver: ${LOG_DRIVER:-local}
restart: always
volumes:
opensearch-data: