From 9393509a5e37e6a1e6a950602d9afb9ac3775a5d Mon Sep 17 00:00:00 2001
From: Vallari Agrawal <vallari.agrawal@ibm.com>
Date: Tue, 22 Oct 2024 20:38:46 +0530
Subject: qa/suite/nvmeof: add asserts to scalability_test.sh

Add assertions to 'status_checks()' function.
Use "apply" and "redeploy", instead of "orch rm" and
"apply" to upscale/downscale gateways.

Signed-off-by: Vallari Agrawal <vallari.agrawal@ibm.com>
---
 .../nvmeof/basic/workloads/nvmeof_scalability.yaml |  5 ++-
 qa/workunits/nvmeof/scalability_test.sh            | 51 +++++++++++++++++-----
 2 files changed, 43 insertions(+), 13 deletions(-)

(limited to 'qa')

diff --git a/qa/suites/nvmeof/basic/workloads/nvmeof_scalability.yaml b/qa/suites/nvmeof/basic/workloads/nvmeof_scalability.yaml
index 12cb50b408d..d66b6fc8093 100644
--- a/qa/suites/nvmeof/basic/workloads/nvmeof_scalability.yaml
+++ b/qa/suites/nvmeof/basic/workloads/nvmeof_scalability.yaml
@@ -31,8 +31,11 @@ tasks:
     no_coverage_and_limits: true
     timeout: 30m
     clients:
-      client.0:
+      client.3:
         - nvmeof/scalability_test.sh nvmeof.a,nvmeof.b
         - nvmeof/scalability_test.sh nvmeof.b,nvmeof.c,nvmeof.d
+        - nvmeof/scalability_test.sh nvmeof.b,nvmeof.c
     env:
       SCALING_DELAYS: '50'
+      RBD_POOL: mypool
+      NVMEOF_GROUP: mygroup0
diff --git a/qa/workunits/nvmeof/scalability_test.sh b/qa/workunits/nvmeof/scalability_test.sh
index 5a26b6284f7..8ede4b7eda2 100755
--- a/qa/workunits/nvmeof/scalability_test.sh
+++ b/qa/workunits/nvmeof/scalability_test.sh
@@ -3,37 +3,64 @@
 
 GATEWAYS=$1 # exmaple "nvmeof.a,nvmeof.b"
 DELAY="${SCALING_DELAYS:-50}"
+POOL="${RBD_POOL:-mypool}"
+GROUP="${NVMEOF_GROUP:-mygroup0}"
+source /etc/ceph/nvmeof.env
 
 if [ -z "$GATEWAYS" ]; then
     echo "At least one gateway needs to be defined for scalability test"
     exit 1
 fi
 
-pip3 install yq
-
 status_checks() {
-    ceph nvme-gw show mypool ''
-    ceph orch ls
-    ceph orch ps 
-    ceph -s
+    expected_count=$1
+
+    output=$(ceph nvme-gw show $POOL $GROUP) 
+    nvme_show=$(echo $output | grep -o '"AVAILABLE"' | wc -l)
+    if [ "$nvme_show" -ne "$expected_count" ]; then
+        return 1
+    fi
+
+    orch_ls=$(ceph orch ls)
+    if ! echo "$orch_ls" | grep -q "$expected_count/$expected_count"; then
+        return 1
+    fi
+
+    output=$(ceph orch ps --service-name nvmeof.$POOL.$GROUP)     
+    orch_ps=$(echo $output | grep -o 'running' | wc -l)
+    if [ "$orch_ps" -ne "$expected_count" ]; then
+        return 1
+    fi
+
+    ceph_status=$(ceph -s)
+    if ! echo "$ceph_status" | grep -q "HEALTH_OK"; then
+        return 1
+    fi
 }
 
+total_gateways_count=$(( $(echo "$NVMEOF_GATEWAY_IP_ADDRESSES" | tr -cd ',' | wc -c) + 1 ))
+scaled_down_gateways_count=$(( total_gateways_count - $(echo "$GATEWAYS" | tr -cd ',' | wc -c) - 1 ))
+
 
 echo "[nvmeof.scale] Setting up config to remove gateways ${GATEWAYS}"
+ceph orch ls --service-name nvmeof.$POOL.$GROUP --export > /tmp/nvmeof-gw.yaml
 ceph orch ls nvmeof --export > /tmp/nvmeof-gw.yaml
 cat /tmp/nvmeof-gw.yaml
-yq "del(.placement.hosts[] | select(. | test(\".*($(echo $GATEWAYS | sed 's/,/|/g'))\")))" /tmp/nvmeof-gw.yaml > /tmp/nvmeof-gw-new.yaml
+
+pattern=$(echo $GATEWAYS | sed 's/,/\\|/g')
+sed "/$pattern/d" /tmp/nvmeof-gw.yaml > /tmp/nvmeof-gw-new.yaml  
 cat /tmp/nvmeof-gw-new.yaml
 
 echo "[nvmeof.scale] Starting scale testing by removing ${GATEWAYS}"
-status_checks
-ceph orch rm nvmeof.mypool && sleep 20 # temp workaround
+status_checks $total_gateways_count 
 ceph orch apply -i /tmp/nvmeof-gw-new.yaml # downscale
+ceph orch redeploy nvmeof.$POOL.$GROUP 
 sleep $DELAY
-status_checks
-ceph orch rm nvmeof.mypool && sleep 20 # temp workaround
+status_checks $scaled_down_gateways_count
+echo "[nvmeof.scale] Downscale complete - removed gateways (${GATEWAYS}); now scaling back up"
 ceph orch apply -i /tmp/nvmeof-gw.yaml #upscale
+ceph orch redeploy nvmeof.$POOL.$GROUP 
 sleep $DELAY
-status_checks
+status_checks $total_gateways_count
 
 echo "[nvmeof.scale] Scale testing passed for ${GATEWAYS}"
-- 
cgit v1.2.3