qa/suites/crimson: Add a test for PG merging in the crimson suite

Signed-off-by: Aishwarya Mathuria <amathuri@redhat.com>

diff --git a/qa/objectstore_crimson/crimson_bluestore.yaml b/qa/objectstore_crimson/crimson_bluestore.yaml
index ff9bcab187d6635459455b642571bdca370854b4..1c5895deed03b386447f0062d8f531ebda626a69 100644 (file)
--- a/qa/objectstore_crimson/crimson_bluestore.yaml
+++ b/qa/objectstore_crimson/crimson_bluestore.yaml
@@ -2,6 +2,8 @@ overrides:
   ceph:
     fs: xfs
     conf:
+      global:
+        crimson_allow_pg_merge: true
       osd:
         # crimson's osd objectstore option
         osd objectstore: bluestore

diff --git a/qa/suites/crimson-rados/singleton/all/osd-pg-merging.yaml b/qa/suites/crimson-rados/singleton/all/osd-pg-merging.yaml
new file mode 100644 (file)
index 0000000..9f4e1a2
--- /dev/null
+++ b/qa/suites/crimson-rados/singleton/all/osd-pg-merging.yaml
@@ -0,0 +1,30 @@
+roles:
+- - mon.a
+  - mon.b
+  - mon.c
+  - mgr.x
+  - osd.0
+  - osd.1
+  - osd.2
+  - client.0
+openstack:
+  - volumes: # attached to each instance
+      count: 3
+      size: 10 # GB
+tasks:
+- install:
+- ceph:
+    pre-mgr-commands:
+      - sudo ceph config set mgr mgr_pool false --force
+    log-ignorelist:
+      - but it is still running
+      - overall HEALTH_
+      - \(PG_
+    conf:
+      osd:
+        osd min pg log entries: 5
+        crimson cpu num: 2
+- workunit:
+    clients:
+      client.0:
+        - rados/test_pg_merging.sh

diff --git a/qa/workunits/rados/test_pg_merging.sh b/qa/workunits/rados/test_pg_merging.sh
new file mode 100755 (executable)
index 0000000..1441511
--- /dev/null
+++ b/qa/workunits/rados/test_pg_merging.sh
@@ -0,0 +1,97 @@
+#!/bin/bash
+
+. $(dirname $0)/../../standalone/ceph-helpers.sh
+
+set -x
+
+function wait_for_merge_completion() {
+    local pool=$1
+    local target=$2
+    local stall_timeout=60
+    local actual=0
+    local last_actual=0
+    local last_progress_time
+
+    wait_for_clean || return 1
+    last_actual=$(ceph pg ls-by-pool $pool --format=json 2>/dev/null | jq -r ".pg_stats | length")
+    last_progress_time=$(date +%s)
+
+    echo "Waiting for pool $pool to reach $target PGs..."
+
+    while true; do
+        actual=$last_actual
+        echo "Current PG count: $actual (Target: $target)"
+        if [ "$actual" -eq "$target" ]; then
+            break
+        fi
+
+        sleep 2
+        wait_for_clean || return 1
+        actual=$(ceph pg ls-by-pool $pool --format=json 2>/dev/null | jq -r ".pg_stats | length")
+
+        if [ "$actual" -lt "$last_actual" ]; then
+            last_progress_time=$(date +%s)
+        fi
+        last_actual=$actual
+
+        if [ $(( $(date +%s) - last_progress_time )) -ge "$stall_timeout" ]; then
+            echo "PG merge made no progress for ${stall_timeout}s in pool $pool (still at $actual PGs)"
+            return 1
+        fi
+    done
+}
+
+function test_pg_merging() {
+    local pool="merge_pool"
+    ceph osd pool delete $pool $pool --yes-i-really-really-mean-it || true
+    create_pool $pool 16
+    wait_for_clean || return 1
+
+    ceph osd pool set $pool nopgchange 0
+    ceph osd pool set $pool crimson_allow_pg_merge true
+
+    # Induce merges
+    ceph osd pool set $pool pg_num 4
+    ceph osd pool set $pool pg_num_min 4
+    
+    wait_for_merge_completion $pool 4 || return 1
+    
+    ceph osd pool set $pool pgp_num 4
+    wait_for_clean || return 1
+}
+
+function test_pg_merging_with_radosbench() {
+    local pool="merge_bench"
+    ceph osd pool delete $pool $pool --yes-i-really-really-mean-it || true
+    create_pool $pool 16
+    wait_for_clean || return 1
+
+    # Start radosbench writes
+    timeout 120 rados bench -p $pool 60 write -b 4096 --no-cleanup &
+    BENCH_PID=$!
+    sleep 5
+
+    # merge
+    ceph osd pool set $pool nopgchange 0
+    ceph osd pool set $pool crimson_allow_pg_merge true
+    ceph osd pool set $pool pg_num 4
+    ceph osd pool set $pool pg_num_min 4
+    
+    # Use the loop to ensure we don't stop until all 12 PGs are merged away
+    wait_for_merge_completion $pool 4 || return 1
+    
+    ceph osd pool set $pool pgp_num 4
+    wait_for_clean || return 1
+
+    # Ensure the bench finished successfully
+    wait $BENCH_PID
+    
+    # Final check
+    actual=$(ceph pg ls-by-pool $pool --format=json 2>/dev/null | jq -r ".pg_stats | length")
+    test "$actual" -eq 4 || return 1
+}
+
+test_pg_merging || { echo "test_pg_merging failed"; exit 1; }
+test_pg_merging_with_radosbench || { echo "test_pg_merging_with_radosbench failed"; exit 1; }
+
+echo "OK"