RAGFlow is an agent-native database: a Retrieval-Augmented Generation (RAG) engine that grew into context infrastructure.
Go agent-harnessagent-native-databaseagentic-aiagentic-retrievalagentic-searchaiai-agentscontext-enginecontext-engineeringcontext-managementharness-engineeringknowledge-compilationragretrieval-augmented-generationretrieval-harnesssearch-harness
Go CI: fix port conflicts (#20157)
Jin Haif1c3a6bbf91a
diff --git a/.github/workflows/sep-tests.yml b/.github/workflows/sep-tests.ymlindex 14c978dc41..0d55407744 100644--- a/.github/workflows/sep-tests.yml+++ b/.github/workflows/sep-tests.yml@@ -604,10 +604,18 @@ jobs:# choosing the same host ports when they land on the same self-hosted runner.# A lock plus reservation file closes the check/start race between parallel jobs.PORT_BASES=(1200 1201 23817 23820 5432 5455 9000 9001 6379 6380 6381 6601 9380 9381 9382 9384 9383 9385 80 443 4222 8222 9900 8123)- PARTITION_SIZE=6000+ PARTITION_SIZE=3000+ # Keep every (base + offset) host port BELOW the Linux ephemeral+ # range (net.ipv4.ip_local_port_range, default 32768-60999). An+ # outbound connection from the host can transiently squat on a port+ # inside that range, and `docker compose up` then fails with+ # "address already in use" even though the reservation succeeded —+ # the reservation cannot see it because `docker ps` only lists+ # container-published ports. Largest base is 23820, so offsets must+ # stay under 8948 for the ceiling to clear 32768.case "${DOC_ENGINE}" in- elasticsearch) PARTITION_BASE=1000 ;;- infinity) PARTITION_BASE=31000 ;;+ elasticsearch) PARTITION_BASE=1000 ;; # offsets 1000-3999 -> ports <= 27819+ infinity) PARTITION_BASE=4000 ;; # offsets 4000-6999 -> ports <= 30819*) echo "Unsupported DOC_ENGINE=${DOC_ENGINE}" >&2; exit 1 ;;esac# Runners run inside docker containers that share the host docker daemon@@ -619,8 +627,8 @@ jobs:# a docker network named ragflow-ci-port-<port> on the shared daemon. docker# network create is NOT idempotent (fails on name conflict), so only one runner# can claim a given port. This is per-PORT, not per-offset, because different- # (base, offset) pairs can produce the same host port (e.g. 6380+36497 ==- # 9380+33497 == 42877); a per-offset guard cannot prevent that collision.+ # (base, offset) pairs can produce the same host port (e.g. 6380+5999 ==+ # 9380+2999 == 12379); a per-offset guard cannot prevent that collision.# - port availability = docker ps published ports (sees every container on the# shared daemon regardless of which runner started it).# No file-system sharing required, no host netns required.@@ -1340,10 +1348,18 @@ jobs:# choosing the same host ports when they land on the same self-hosted runner.# A lock plus reservation file closes the check/start race between parallel jobs.PORT_BASES=(1200 1201 23817 23820 5432 5455 9000 9001 6379 6380 6381 6601 9380 9381 9382 9384 9383 9385 80 443 4222 8222 9900 8123)- PARTITION_SIZE=6000+ PARTITION_SIZE=3000+ # Keep every (base + offset) host port BELOW the Linux ephemeral+ # range (net.ipv4.ip_local_port_range, default 32768-60999). An+ # outbound connection from the host can transiently squat on a port+ # inside that range, and `docker compose up` then fails with+ # "address already in use" even though the reservation succeeded —+ # the reservation cannot see it because `docker ps` only lists+ # container-published ports. Largest base is 23820, so offsets must+ # stay under 8948 for the ceiling to clear 32768.case "${DOC_ENGINE}" in- elasticsearch) PARTITION_BASE=1000 ;;- infinity) PARTITION_BASE=31000 ;;+ elasticsearch) PARTITION_BASE=1000 ;; # offsets 1000-3999 -> ports <= 27819+ infinity) PARTITION_BASE=4000 ;; # offsets 4000-6999 -> ports <= 30819*) echo "Unsupported DOC_ENGINE=${DOC_ENGINE}" >&2; exit 1 ;;esac# Runners run inside docker containers that share the host docker daemon@@ -1355,8 +1371,8 @@ jobs:# a docker network named ragflow-ci-port-<port> on the shared daemon. docker# network create is NOT idempotent (fails on name conflict), so only one runner# can claim a given port. This is per-PORT, not per-offset, because different- # (base, offset) pairs can produce the same host port (e.g. 6380+36497 ==- # 9380+33497 == 42877); a per-offset guard cannot prevent that collision.+ # (base, offset) pairs can produce the same host port (e.g. 6380+5999 ==+ # 9380+2999 == 12379); a per-offset guard cannot prevent that collision.# - port availability = docker ps published ports (sees every container on the# shared daemon regardless of which runner started it).# No file-system sharing required, no host netns required.@@ -2012,10 +2028,18 @@ jobs:# choosing the same host ports when they land on the same self-hosted runner.# A lock plus reservation file closes the check/start race between parallel jobs.PORT_BASES=(1200 1201 23817 23820 5432 5455 9000 9001 6379 6380 6381 6601 9380 9381 9382 9384 9383 9385 80 443 4222 8222 9900 8123)- PARTITION_SIZE=6000+ PARTITION_SIZE=3000+ # Keep every (base + offset) host port BELOW the Linux ephemeral+ # range (net.ipv4.ip_local_port_range, default 32768-60999). An+ # outbound connection from the host can transiently squat on a port+ # inside that range, and `docker compose up` then fails with+ # "address already in use" even though the reservation succeeded —+ # the reservation cannot see it because `docker ps` only lists+ # container-published ports. Largest base is 23820, so offsets must+ # stay under 8948 for the ceiling to clear 32768.case "${DOC_ENGINE}" in- elasticsearch) PARTITION_BASE=1000 ;;- infinity) PARTITION_BASE=31000 ;;+ elasticsearch) PARTITION_BASE=1000 ;; # offsets 1000-3999 -> ports <= 27819+ infinity) PARTITION_BASE=4000 ;; # offsets 4000-6999 -> ports <= 30819*) echo "Unsupported DOC_ENGINE=${DOC_ENGINE}" >&2; exit 1 ;;esac# Runners run inside docker containers that share the host docker daemon@@ -2027,8 +2051,8 @@ jobs:# a docker network named ragflow-ci-port-<port> on the shared daemon. docker# network create is NOT idempotent (fails on name conflict), so only one runner# can claim a given port. This is per-PORT, not per-offset, because different- # (base, offset) pairs can produce the same host port (e.g. 6380+36497 ==- # 9380+33497 == 42877); a per-offset guard cannot prevent that collision.+ # (base, offset) pairs can produce the same host port (e.g. 6380+5999 ==+ # 9380+2999 == 12379); a per-offset guard cannot prevent that collision.# - port availability = docker ps published ports (sees every container on the# shared daemon regardless of which runner started it).# No file-system sharing required, no host netns required.