FDB-683: Ceph/RADOS backend

.github/ci-config.yml

@@ -2,7 +2,7 @@ dependencies: |
   ecmwf/ecbuild
   Deutsches-Klimarechenzentrum/libaec@refs/tags/v1.1.6
   ecmwf/eccodes
-  ecmwf/eckit
+  ecmwf/eckit@feature/ECKIT-684-ceph-backend
   ecmwf/metkit
 dependency_branch: develop
 cmake_options: -DENABLE_MEMFS=ON -DENABLE_DUMMY_DAOS=ON -DENABLE_FAMFDB=ON

.github/ci-hpc-config.yml

@@ -4,7 +4,7 @@ build:
   dependencies:
     - ecmwf/ecbuild@develop
     - ecmwf/eccodes@develop
-    - ecmwf/eckit@develop
+    - ecmwf/eckit@feature/ECKIT-684-ceph-backend
     - ecmwf/metkit@develop
   cmake_options:
     - -DENABLE_LUSTRE=OFF

.github/workflows/ci-rados.yml

@@ -0,0 +1,253 @@
+name: ci-rados
+
+on:
+  push:
+    branches:
+      - "master"
+      - "develop"
+    tags-ignore:
+      - "**"
+
+  # Trigger the workflow on pull request
+  pull_request: ~
+
+  # Trigger on public pull request approval
+  pull_request_target:
+    types: [labeled]
+
+  # Trigger the workflow manually
+  workflow_dispatch: ~
+
+# avoid duplicate runs
+concurrency:
+  group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  rados:
+    name: Test Ceph/RADOS Backend
+    runs-on: ubuntu-latest
+
+    env:
+      CEPH_ETC: /etc/ceph
+      ECKIT_RADOS_CLUSTER_NAME: ceph
+      ECKIT_RADOS_CLUSTER_USER: client.admin
+      FDB_RADOS_TEST_POOL: fdb_test
+      # Install prefix shared by all bundle projects (eckit, eccodes, metkit, fdb).
+      INSTALL_PREFIX: ${{ github.workspace }}/install
+      # Local dirs used to cache apt archives and the Ceph docker image.
+      APT_CACHE: ${{ github.workspace }}/.apt-cache
+      CEPH_IMAGE: quay.io/ceph/demo@sha256:522483cf07cfce6386b8e18a3edfa88a1b32c688dee231d0a6962b1513557723
+      CEPH_IMAGE_CACHE: ${{ github.workspace }}/.docker-ceph
+      CMAKE_FLAGS: -DENABLE_AEC=OFF -DENABLE_EXAMPLES=OFF -DENABLE_EXPERIMENTAL=OFF -DENABLE_NETCDF=OFF
+
+    steps:
+      - name: Checkout ecbuild
+        uses: actions/checkout@v4
+        with:
+          repository: ecmwf/ecbuild
+          path: ecbuild
+
+      - name: Checkout eckit
+        uses: actions/checkout@v4
+        with:
+          repository: ecmwf/eckit
+          ref: feature/ECKIT-684-ceph-backend
+          path: eckit
+
+      - name: Checkout eccodes
+        uses: actions/checkout@v4
+        with:
+          repository: ecmwf/eccodes
+          path: eccodes
+
+      - name: Checkout metkit
+        uses: actions/checkout@v4
+        with:
+          repository: ecmwf/metkit
+          path: metkit
+
+      - name: Checkout fdb
+        uses: actions/checkout@v4
+        with:
+          path: fdb
+
+      - name: Cache apt packages
+        uses: actions/cache@v4
+        with:
+          path: ${{ github.workspace }}/.apt-cache
+          # Bump the suffix to invalidate when the package list changes.
+          key: apt-rados-${{ runner.os }}-v2
+
+      - name: Install build dependencies
+        run: |
+          mkdir -p "${APT_CACHE}/archives/partial"
+          sudo apt-get update
+          sudo apt-get install -y --no-install-recommends \
+            -o Dir::Cache::archives="${APT_CACHE}/archives" \
+            -o APT::Keep-Downloaded-Packages=true \
+            build-essential \
+            cmake \
+            jq \
+            ninja-build \
+            gfortran \
+            libopenmpi-dev \
+            libssl-dev \
+            uuid-dev \
+            libbz2-dev \
+            libcurl4-openssl-dev \
+            libaec-dev \
+            librados-dev
+          # Make the cached .deb archives readable/writable by the runner user.
+          sudo chown -R "$(id -u):$(id -g)" "${APT_CACHE}"
+
+      - name: Cache Ceph image
+        uses: actions/cache@v4
+        with:
+          path: ${{ github.workspace }}/.docker-ceph
+          # Bump the suffix to refresh the pinned image snapshot.
+          key: ceph-image-522483cf07cf-v2
+
+      - name: Load or pull Ceph image
+        run: |
+          if [ -f "${CEPH_IMAGE_CACHE}/ceph-demo.tar" ]; then
+            docker load -i "${CEPH_IMAGE_CACHE}/ceph-demo.tar"
+          else
+            docker pull "${CEPH_IMAGE}"
+            mkdir -p "${CEPH_IMAGE_CACHE}"
+            docker save "${CEPH_IMAGE}" -o "${CEPH_IMAGE_CACHE}/ceph-demo.tar"
+          fi
+
+      - name: Start Ceph cluster
+        run: |
+          sudo mkdir -p "${CEPH_ETC}"
+          docker run -d --name ceph-demo \
+            --network host \
+            -e MON_IP=127.0.0.1 \
+            -e CEPH_PUBLIC_NETWORK=0.0.0.0/0 \
+            -e CEPH_DEMO_UID=ci \
+            -e DEMO_DAEMONS=mon,mgr,osd \
+            -v "${CEPH_ETC}:/etc/ceph" \
+            "${CEPH_IMAGE}" demo
+
+      - name: Wait for Ceph and create test pool
+        run: |
+          echo "Waiting for Ceph to become ready..."
+          for i in $(seq 1 60); do
+            if docker exec ceph-demo ceph osd pool ls >/dev/null 2>&1; then
+              ready=1
+              break
+            fi
+            sleep 5
+          done
+          if [ "${ready:-0}" != "1" ]; then
+            echo "Ceph did not become ready in time" >&2
+            docker logs ceph-demo || true
+            exit 1
+          fi
+          echo "Waiting for Ceph cluster health..."
+          for i in $(seq 1 60); do
+            if docker exec ceph-demo ceph health 2>/dev/null | grep -q '^HEALTH_OK'; then
+              health_ready=1
+              break
+            fi
+            sleep 5
+          done
+          if [ "${health_ready:-0}" != "1" ]; then
+            echo "Ceph cluster did not become healthy in time" >&2
+            docker exec ceph-demo ceph status || true
+            docker logs ceph-demo || true
+            exit 1
+          fi
+          docker exec ceph-demo ceph config set mon mon_allow_pool_size_one true
+          docker exec ceph-demo ceph osd pool create "${FDB_RADOS_TEST_POOL}" 8 8
+          # The demo cluster has one OSD; its test pool must use a single replica to become active+clean.
+          docker exec ceph-demo ceph osd pool set "${FDB_RADOS_TEST_POOL}" size 1 --yes-i-really-mean-it
+          docker exec ceph-demo ceph osd pool set "${FDB_RADOS_TEST_POOL}" min_size 1
+          docker exec ceph-demo ceph osd pool application enable "${FDB_RADOS_TEST_POOL}" rados
+          echo "Waiting for ${FDB_RADOS_TEST_POOL} placement groups to become active+clean..."
+          for i in $(seq 1 60); do
+            if docker exec ceph-demo ceph pg stat 2>/dev/null | \
+              awk '$2 == "pgs:" { gsub("[,;]", "", $4); exit !($1 == $3 && $4 == "active+clean") }'; then
+              pool_ready=1
+              break
+            fi
+            sleep 5
+          done
+          if [ "${pool_ready:-0}" != "1" ]; then
+            echo "Ceph test pool did not become active+clean in time" >&2
+            docker exec ceph-demo ceph status || true
+            docker exec ceph-demo ceph pg stat || true
+            exit 1
+          fi
+          sudo chmod 0644 "${CEPH_ETC}/ceph.client.admin.keyring"
+
+      - name: Build eckit
+        run: |
+          # eckit locates ecbuild via find_package(ecbuild ... HINTS .../../ecbuild),
+          # which resolves to the sibling ecbuild checkout above.
+          cmake -S eckit -B build/eckit -G Ninja \
+            -DCMAKE_BUILD_TYPE=Release \
+            -DCMAKE_INSTALL_PREFIX="${INSTALL_PREFIX}" \
+            ${CMAKE_FLAGS} \
+            -DENABLE_MPI=ON \
+            -DENABLE_RADOS=ON \
+            -DENABLE_RADOS_TESTS_MANAGE_POOLS=OFF
+          cmake --build build/eckit --parallel
+          cmake --install build/eckit
+
+      - name: Build eccodes
+        run: |
+          cmake -S eccodes -B build/eccodes -G Ninja \
+            -DCMAKE_BUILD_TYPE=Release \
+            -DCMAKE_INSTALL_PREFIX="${INSTALL_PREFIX}" \
+            -DCMAKE_PREFIX_PATH="${INSTALL_PREFIX}" \
+            ${CMAKE_FLAGS} \
+            -DENABLE_JPG=OFF \
+            -DENABLE_FORTRAN=ON \
+            -DENABLE_MEMFS=ON \
+            -DENABLE_ECCODES_THREADS=ON
+          cmake --build build/eccodes --parallel
+          cmake --install build/eccodes
+
+      - name: Build metkit
+        run: |
+          cmake -S metkit -B build/metkit -G Ninja \
+            -DCMAKE_BUILD_TYPE=Release \
+            -DCMAKE_INSTALL_PREFIX="${INSTALL_PREFIX}" \
+            -DCMAKE_PREFIX_PATH="${INSTALL_PREFIX}" \
+            ${CMAKE_FLAGS}
+          cmake --build build/metkit --parallel
+          cmake --install build/metkit
+
+      - name: Build fdb
+        run: |
+          cmake -S fdb -B build/fdb -G Ninja \
+            -DCMAKE_BUILD_TYPE=Release \
+            -DCMAKE_INSTALL_PREFIX="${INSTALL_PREFIX}" \
+            -DCMAKE_PREFIX_PATH="${INSTALL_PREFIX}" \
+            ${CMAKE_FLAGS} \
+            -DENABLE_RADOSFDB=ON \
+            -DENABLE_RADOS_TESTS_MANAGE_POOLS=OFF \
+            -DFDB_RADOS_TEST_POOL="${FDB_RADOS_TEST_POOL}"
+          cmake --build build/fdb --parallel
+
+      - name: Run RADOS FDB tests
+        env:
+          ECKIT_RADOS_CLUSTER_CONF: /etc/ceph/ceph.conf
+        run: |
+          ctest --test-dir build/fdb -L rados --output-on-failure
+
+      - name: Dump Ceph diagnostics on failure
+        if: failure()
+        run: |
+          echo "=== Ceph status ==="
+          docker exec ceph-demo ceph status || true
+          echo "=== Ceph capacity ==="
+          docker exec ceph-demo ceph df || true
+          echo "=== Test-pool objects ==="
+          docker exec ceph-demo rados -p "${FDB_RADOS_TEST_POOL}" ls || true
+          echo "=== CTest temporary files ==="
+          ls -la build/fdb/Testing/Temporary/ || true
+          echo "=== Ceph container logs ==="
+          docker logs ceph-demo || true

.gitignore

@@ -20,3 +20,4 @@ __pycache__/
 # Rust
 rust/target/
 rust/Cargo.lock
+doc-build

CMakeLists.txt

@@ -49,11 +49,16 @@ ecbuild_find_package( NAME metkit VERSION 1.15 REQUIRED )
 ### FDB backend in CEPH object store (using Rados)
 
 find_package( RADOS QUIET )
+
 ecbuild_add_option( FEATURE RADOSFDB  # option defined in fdb5_config.h
                     CONDITION eckit_HAVE_RADOS AND RADOS_FOUND
                     DEFAULT OFF
                     DESCRIPTION "Ceph/Rados support for FDB Store" )
 
+ecbuild_add_option( FEATURE RADOS_TESTS_MANAGE_POOLS
+                    DEFAULT OFF
+                    DESCRIPTION "Have unit tests create pools automatically rather than using an existing pool specified in the FDB_RADOS_TEST_POOL cmake variable." )
+
 ### FDB backend in indexed filesystem with table-of-contents, i.e. TOC
 ### Supports Lustre parallel filesystem stripping control
 

cmake/FindRADOS.cmake

@@ -1,49 +1,51 @@
-# (C) Copyright 2011- ECMWF.
+# (C) Copyright 2026- ECMWF.
 #
 # This software is licensed under the terms of the Apache Licence Version 2.0
 # which can be obtained at http://www.apache.org/licenses/LICENSE-2.0.
 # In applying this licence, ECMWF does not waive the privileges and immunities
 # granted to it by virtue of its status as an intergovernmental organisation nor
 # does it submit to any jurisdiction.
-
-# - Try to find Rados
-# Once done this will define
-#
-#  RADOS_FOUND         - system has Armadillo
-#  RADOS_INCLUDE_DIRS  - the Armadillo include directory
-#  RADOS_LIBRARIES     - the Armadillo library
-#  RADOS_VERSION - This is set to $major.$minor.$patch (eg. 0.9.8)
 #
-# The following paths will be searched with priority if set in CMake or env
+# This module defines the following variables:
+#  RADOS_INCLUDE_DIRS - Where to find rados/librados.h
+#  RADOS_LIBRARIES    - The libraries needed to use Rados
+#  RADOS_FOUND        - True if Rados was found
 #
-#  RADOS_PATH          - prefix path of the Armadillo installation
-#  RADOS_ROOT              - Set this variable to the root installation
-
-# Search with priority for RADOS_PATH if given as CMake or env var
-
-find_path(RADOS_INCLUDE_DIR rados/librados.hpp
-          HINTS $ENV{RADOS_ROOT} ${RADOS_ROOT}
-          PATHS ${RADOS_PATH} ENV RADOS_PATH
-          PATH_SUFFIXES include NO_DEFAULT_PATH)
-
-find_path(RADOS_INCLUDE_DIR rados/librados.hpp PATH_SUFFIXES include )
-
-# Search with priority for RADOS_PATH if given as CMake or env var
-find_library(RADOS_LIBRARY rados
-            HINTS $ENV{RADOS_ROOT} ${RADOS_ROOT}
-            PATHS ${RADOS_PATH} ENV RADOS_PATH
-            PATH_SUFFIXES lib64 lib NO_DEFAULT_PATH)
-
-find_library( RADOS_LIBRARY rados PATH_SUFFIXES lib64 lib )
-
-set( RADOS_LIBRARIES    ${RADOS_LIBRARY} )
-set( RADOS_INCLUDE_DIRS ${RADOS_INCLUDE_DIR} )
-
+# This module also defines the following IMPORTED target:
+#  Ceph::RADOS
+
+# Find the header path by looking for the subdirectory file
+find_path(RADOS_INCLUDE_DIR
+    NAMES rados/librados.h
+    DOC "Path to Rados include directory"
+)
+
+# Find the library
+find_library(RADOS_LIBRARY
+    NAMES rados
+    DOC "Path to Rados library"
+)
+
+# Handle the QUIETLY and REQUIRED arguments and set RADOS_FOUND to TRUE if
+# all listed variables are TRUE.
 include(FindPackageHandleStandardArgs)
-
-# handle the QUIET and REQUIRED arguments and set RADOS_FOUND to TRUE
-# if all listed variables are TRUE
-# Note: capitalisation of the package name must be the same as in the file name
-find_package_handle_standard_args(RADOS DEFAULT_MSG RADOS_LIBRARY RADOS_INCLUDE_DIR)
-
+find_package_handle_standard_args(RADOS
+    REQUIRED_VARS RADOS_LIBRARY RADOS_INCLUDE_DIR
+)
+
+if(RADOS_FOUND)
+    set(RADOS_LIBRARIES ${RADOS_LIBRARY})
+    set(RADOS_INCLUDE_DIRS ${RADOS_INCLUDE_DIR})
+
+    # Create an modern generic imported target
+    if(NOT TARGET Ceph::RADOS)
+        add_library(Ceph::RADOS UNKNOWN IMPORTED)
+        set_target_properties(Ceph::RADOS PROPERTIES
+            INTERFACE_INCLUDE_DIRECTORIES "${RADOS_INCLUDE_DIRS}"
+            IMPORTED_LOCATION "${RADOS_LIBRARY}"
+        )
+    endif()
+endif()
+
+# Hide these variables from the GUI cache view
 mark_as_advanced(RADOS_INCLUDE_DIR RADOS_LIBRARY)