[TFA] tier-2_cephfs_test-metrics suite failures

Signed-off-by: Amarnath K <[email protected]>

Author: Amarnath K

tests/cephfs/cephfs_metrics/cephfs_metrics_scale.py

@@ -1,11 +1,14 @@
+import json
 import random
 import string
 import time
 import traceback
+from json import JSONDecodeError
 
 from ceph.ceph import CommandFailed
 from tests.cephfs.cephfs_utilsV1 import FsUtils
 from utility.log import Log
+from utility.retry import retry
 
 log = Log(__name__)
 
@@ -24,6 +27,109 @@ class Metrics_Value_Not_Matching(Exception):
     pass
 
 
+@retry((JSONDecodeError, CommandFailed), tries=5, delay=5)
+def get_client_id(client, rank=0, mounted_dir="", fs_name="cephfs"):
+    ranked_mds, _ = client.exec_command(
+        sudo=True,
+        cmd=f"ceph fs status {fs_name} -f json | jq '.mdsmap[] | select(.rank == {rank}) | .name'",
+    )
+    log.info("Executing MDS name with rank command: %s", ranked_mds)
+    ranked_mds = ranked_mds.replace('"', "").replace("\n", "")
+    client_id_cmd = (
+        f"ceph tell mds.{ranked_mds} session ls | jq '.[] | select(.client_metadata.mount_point"
+        f' != null and (.client_metadata.mount_point | contains("{mounted_dir}"))) | .id\''
+    )
+    log.info(f"Executing Client ID Command : {client_id_cmd}")
+    client_id, _ = client.exec_command(sudo=True, cmd=client_id_cmd)
+    client_id = client_id.replace('"', "").replace("\n", "")
+    if client_id == "":
+        log.error(f"Client not found for Mounted Directory : {mounted_dir}")
+        raise CommandFailed(f"Client not found for Mounted Directory : {mounted_dir}")
+    log.info(f"Client ID :[{client_id}] for Mounted Directory : [{mounted_dir}]")
+    return client_id, rank
+
+
+@retry((JSONDecodeError, CommandFailed), tries=5, delay=10)
+def get_mds_metrics_for_client(
+    client, client_id, rank, mds_rank=0, mounted_dir="", fs_name="cephfs"
+):
+    ranked_mds, _ = client.exec_command(
+        sudo=True,
+        cmd=f"ceph fs status {fs_name} -f json | jq '.mdsmap[] | select(.rank == {mds_rank}) | .name'",
+    )
+    log.info(f"Executing MDS name with rank command: {ranked_mds}")
+    ranked_mds = ranked_mds.replace('"', "").replace("\n", "")
+    log.info(f"Client ID :[{client_id}] for Mounted Directory : [{mounted_dir}]")
+    cmd = f""" ceph tell mds.{ranked_mds} counter dump 2>/dev/null | \
+                 jq -r '. | to_entries | map(select(.key | match("mds_client_metrics"))) | \
+                 .[].value[] | select(.labels.client != null and (.labels.client | contains("{client_id}"))
+                 and (.labels.rank == "{rank}"))'
+                 """
+    metrics_out, _ = client.exec_command(sudo=True, cmd=cmd)
+    log.info(
+        f"Metrics for MDS : {ranked_mds} Mounted Directory: {mounted_dir} and Client : {client_id} is {metrics_out}"
+    )
+    if metrics_out == "":
+        log.error(f"Metrics not found for MDS : {ranked_mds}")
+        raise CommandFailed(f"Client not found for Mounted Directory : {mounted_dir}")
+    metrics_out = json.loads(str(metrics_out))
+    return metrics_out
+
+
+def get_mds_metrics_from_ranks(ranks, fs_util, client, mount_dir, cephfs):
+    """
+    Try fetching MDS metrics for the given client and mount_dir from the list of ranks.
+
+    Returns:
+        dict: MDS metrics if found.
+
+    Raises:
+        CommandFailed: If no metrics are found from any rank.
+    """
+    client_id = None
+    client_rank = None
+
+    # Step 1: Get client_id from one of the MDS ranks
+    for rank in ranks:
+        try:
+            client_id, client_rank = get_client_id(client, rank, mount_dir, cephfs)
+            if client_id:
+                log.info(f"Found client ID '{client_id}' from rank {client_rank}")
+                break
+        except Exception as e:
+            log.warning(
+                f"Rank {rank}: Failed to get client ID for mount {mount_dir}: {e}"
+            )
+            continue
+
+    if not client_id:
+        raise CommandFailed(f"Client not found in any MDS ranks for mount {mount_dir}")
+
+    # Step 2: Use client_id and try to collect metrics from all MDS ranks
+    for rank in ranks:
+        try:
+            mds_metric = get_mds_metrics_for_client(
+                client,
+                client_id,
+                client_rank,
+                mds_rank=rank,
+                mounted_dir=mount_dir,
+                fs_name=cephfs,
+            )
+            if mds_metric and mds_metric != 1:
+                log.info(f"Successfully got MDS metrics from rank {rank}")
+                return mds_metric
+        except Exception as e:
+            log.warning(
+                f"Rank {rank}: Failed to fetch metrics for client {client_id}: {e}"
+            )
+            continue
+
+    raise CommandFailed(
+        f"Metrics not found for client {client_id} in any of the MDS ranks for mount {mount_dir}"
+    )
+
+
 def run(ceph_cluster, **kw):
     try:
         tc = "CEPH-83588355"
@@ -104,17 +210,18 @@ def run(ceph_cluster, **kw):
         fs_util.fuse_mount([client4], fuse_mounting_dir_4)
         # Get initial MDS metrics
         # pdb.set_trace()
-        mds_metric_client1 = fs_util.get_mds_metrics(
-            client1, 0, fuse_mounting_dir_1, cephfs
+
+        mds_metric_client1 = get_mds_metrics_from_ranks(
+            [0, 1, 2], fs_util, client1, fuse_mounting_dir_1, cephfs
         )
-        mds_metric_client2 = fs_util.get_mds_metrics(
-            client2, 0, fuse_mounting_dir_2, cephfs
+        mds_metric_client2 = get_mds_metrics_from_ranks(
+            [0, 1, 2], fs_util, client2, fuse_mounting_dir_2, cephfs
         )
-        mds_metric_client3 = fs_util.get_mds_metrics(
-            client3, 0, fuse_mounting_dir_3, cephfs
+        mds_metric_client3 = get_mds_metrics_from_ranks(
+            [0, 1, 2], fs_util, client3, fuse_mounting_dir_3, cephfs
         )
-        mds_metric_client4 = fs_util.get_mds_metrics(
-            client4, 0, fuse_mounting_dir_4, cephfs
+        mds_metric_client4 = get_mds_metrics_from_ranks(
+            [0, 1, 2], fs_util, client4, fuse_mounting_dir_4, cephfs
         )
 
         log.info(f"mds_metric_client1: {mds_metric_client1}")
@@ -132,11 +239,11 @@ def run(ceph_cluster, **kw):
         inode_list = ["opened_inodes", "pinned_icaps", "total_inodes"]
 
         # Get initial inode metrics for client1 and client3
-        client1_pre_inode_metrics = fs_util.get_mds_metrics(
-            client1, 0, fuse_mounting_dir_1, cephfs
+        client1_pre_inode_metrics = get_mds_metrics_from_ranks(
+            [0, 1, 2], fs_util, client1, fuse_mounting_dir_1, cephfs
         )
-        client3_pre_inode_metrics = fs_util.get_mds_metrics(
-            client3, 0, fuse_mounting_dir_3, cephfs
+        client3_pre_inode_metrics = get_mds_metrics_from_ranks(
+            [0, 1, 2], fs_util, client3, fuse_mounting_dir_3, cephfs
         )
         log.info(f"client1_pre_inode_metrics: {client1_pre_inode_metrics}")
         log.info(f"client3_pre_inode_metrics: {client3_pre_inode_metrics}")
@@ -165,11 +272,11 @@ def run(ceph_cluster, **kw):
 
         log.info("Writing files is done for client1 and client3")
         log.info("Get metrics only for client1 and client3")
-        client1_post_inode_metrics = fs_util.get_mds_metrics(
-            client1, 0, fuse_mounting_dir_1, cephfs
+        client1_post_inode_metrics = get_mds_metrics_from_ranks(
+            [0, 1, 2], fs_util, client1, fuse_mounting_dir_1, cephfs
         )
-        client3_post_inode_metrics = fs_util.get_mds_metrics(
-            client3, 0, fuse_mounting_dir_3, cephfs
+        client3_post_inode_metrics = get_mds_metrics_from_ranks(
+            [0, 1, 2], fs_util, client3, fuse_mounting_dir_3, cephfs
         )
         log.info(f"client1_post_inode_metrics: {client1_post_inode_metrics}")
         log.info(f"client3_post_inode_metrics: {client3_post_inode_metrics}")
@@ -197,11 +304,11 @@ def run(ceph_cluster, **kw):
         )
         file_paths_client2 = []
         file_paths_client4 = []
-        pre_opened_files_client2 = fs_util.get_mds_metrics(
-            client2, 0, fuse_mounting_dir_2, cephfs
+        pre_opened_files_client2 = get_mds_metrics_from_ranks(
+            [0, 1, 2], fs_util, client2, fuse_mounting_dir_2, cephfs
         )["counters"]["opened_files"]
-        pre_opened_files_client4 = fs_util.get_mds_metrics(
-            client4, 0, fuse_mounting_dir_4, cephfs
+        pre_opened_files_client4 = get_mds_metrics_from_ranks(
+            [0, 1, 2], fs_util, client4, fuse_mounting_dir_4, cephfs
         )["counters"]["opened_files"]
         log.info(f"pre_opened_files_client2: {pre_opened_files_client2}")
         log.info(f"pre_opened_files_client4: {pre_opened_files_client4}")
@@ -240,11 +347,11 @@ def run(ceph_cluster, **kw):
         log.info(f"Number of PID4s from opening files: {pids4}")
         time.sleep(10)
         log.info("Get final MDS metrics after opening files")
-        client2_post_opened_files = fs_util.get_mds_metrics(
-            client2, 0, fuse_mounting_dir_2, cephfs
+        client2_post_opened_files = get_mds_metrics_from_ranks(
+            [0, 1, 2], fs_util, client2, fuse_mounting_dir_2, cephfs
         )["counters"]["opened_files"]
-        client4_post_opened_files = fs_util.get_mds_metrics(
-            client4, 0, fuse_mounting_dir_4, cephfs
+        client4_post_opened_files = get_mds_metrics_from_ranks(
+            [0, 1, 2], fs_util, client4, fuse_mounting_dir_4, cephfs
         )["counters"]["opened_files"]
         log.info(f"client2_post_opened_files: {client2_post_opened_files}")
         log.info(f"client4_post_opened_files: {client4_post_opened_files}")
@@ -268,11 +375,11 @@ def run(ceph_cluster, **kw):
             log.error(f"Failed to kill tail processes: {e}")
         time.sleep(5)
         log.info("Get final MDS metrics after killing the PIDs")
-        post_opened_files_client2 = fs_util.get_mds_metrics(
-            client2, 0, fuse_mounting_dir_2, cephfs
+        post_opened_files_client2 = get_mds_metrics_from_ranks(
+            [0, 1, 2], fs_util, client2, fuse_mounting_dir_2, cephfs
         )["counters"]["opened_files"]
-        post_opened_files_client4 = fs_util.get_mds_metrics(
-            client4, 0, fuse_mounting_dir_4, cephfs
+        post_opened_files_client4 = get_mds_metrics_from_ranks(
+            [0, 1, 2], fs_util, client4, fuse_mounting_dir_4, cephfs
         )["counters"]["opened_files"]
         log.info(f"post_opened_files_client2: {post_opened_files_client2}")
         log.info(f"post_opened_files_client4: {post_opened_files_client4}")
@@ -288,11 +395,11 @@ def run(ceph_cluster, **kw):
                 "Failed to verify opened_files for client4"
             )
         log.info("Verify if other clients opened_files metrics remain same")
-        post_opened_files_client1 = fs_util.get_mds_metrics(
-            client1, 0, fuse_mounting_dir_1, cephfs
+        post_opened_files_client1 = get_mds_metrics_from_ranks(
+            [0, 1, 2], fs_util, client1, fuse_mounting_dir_1, cephfs
         )["counters"]["opened_files"]
-        post_opened_files_client3 = fs_util.get_mds_metrics(
-            client3, 0, fuse_mounting_dir_3, cephfs
+        post_opened_files_client3 = get_mds_metrics_from_ranks(
+            [0, 1, 2], fs_util, client3, fuse_mounting_dir_3, cephfs
         )["counters"]["opened_files"]
         log.info(f"post_opened_files_client1: {post_opened_files_client1}")
         log.info(f"post_opened_files_client3: {post_opened_files_client3}")
@@ -315,17 +422,17 @@ def run(ceph_cluster, **kw):
         log.info(
             "Increase only Client2 and Client4 dentry metrics and other clients should remain same"
         )
-        pre_dentry_client1 = fs_util.get_mds_metrics(
-            client1, 0, fuse_mounting_dir_1, cephfs
+        pre_dentry_client1 = get_mds_metrics_from_ranks(
+            [0, 1, 2], fs_util, client1, fuse_mounting_dir_1, cephfs
         )["counters"]
-        pre_dentry_client2 = fs_util.get_mds_metrics(
-            client2, 0, fuse_mounting_dir_2, cephfs
+        pre_dentry_client2 = get_mds_metrics_from_ranks(
+            [0, 1, 2], fs_util, client2, fuse_mounting_dir_2, cephfs
         )["counters"]
-        pre_dentry_client3 = fs_util.get_mds_metrics(
-            client3, 0, fuse_mounting_dir_3, cephfs
+        pre_dentry_client3 = get_mds_metrics_from_ranks(
+            [0, 1, 2], fs_util, client3, fuse_mounting_dir_3, cephfs
         )["counters"]
-        pre_dentry_client4 = fs_util.get_mds_metrics(
-            client4, 0, fuse_mounting_dir_4, cephfs
+        pre_dentry_client4 = get_mds_metrics_from_ranks(
+            [0, 1, 2], fs_util, client4, fuse_mounting_dir_4, cephfs
         )["counters"]
         log.info(f"pre_dentry_client2: {pre_dentry_client2}")
         log.info(f"pre_dentry_client4: {pre_dentry_client4}")
@@ -368,11 +475,11 @@ def run(ceph_cluster, **kw):
             client4.exec_command(sudo=True, cmd=f"ls {fuse_mounting_dir_4}/{dir}/")
         time.sleep(5)
         log.info("Get final MDS metrics after creating directories and files")
-        post_dentry_client2 = fs_util.get_mds_metrics(
-            client2, 0, fuse_mounting_dir_2, cephfs
+        post_dentry_client2 = get_mds_metrics_from_ranks(
+            [0, 1, 2], fs_util, client2, fuse_mounting_dir_2, cephfs
         )["counters"]
-        post_dentry_client4 = fs_util.get_mds_metrics(
-            client4, 0, fuse_mounting_dir_4, cephfs
+        post_dentry_client4 = get_mds_metrics_from_ranks(
+            [0, 1, 2], fs_util, client4, fuse_mounting_dir_4, cephfs
         )["counters"]
         log.info(f"post_dentry_client2: {post_dentry_client2}")
         log.info(f"post_dentry_client4: {post_dentry_client4}")
@@ -388,11 +495,11 @@ def run(ceph_cluster, **kw):
                     f"Failed to verify {dentry} for client4"
                 )
         log.info("Verify if other clients dentry metrics remain same")
-        post_dentry_client1 = fs_util.get_mds_metrics(
-            client1, 0, fuse_mounting_dir_1, cephfs
+        post_dentry_client1 = get_mds_metrics_from_ranks(
+            [0, 1, 2], fs_util, client1, fuse_mounting_dir_1, cephfs
         )["counters"]
-        post_dentry_client3 = fs_util.get_mds_metrics(
-            client3, 0, fuse_mounting_dir_3, cephfs
+        post_dentry_client3 = get_mds_metrics_from_ranks(
+            [0, 1, 2], fs_util, client3, fuse_mounting_dir_3, cephfs
         )["counters"]
         log.info(f"post_dentry_client1: {post_dentry_client1}")
         log.info(f"post_dentry_client3: {post_dentry_client3}")
@@ -415,11 +522,11 @@ def run(ceph_cluster, **kw):
         )
         # Using scp from client1 to client2, it will increase total_read_ops and total_read_size in client1
         # In client2, it will increase total_write_ops and total_write_size
-        pre_read_ops_client1 = fs_util.get_mds_metrics(
-            client1, 0, fuse_mounting_dir_1, cephfs
+        pre_read_ops_client1 = get_mds_metrics_from_ranks(
+            [0, 1, 2], fs_util, client1, fuse_mounting_dir_1, cephfs
         )["counters"]
-        pre_write_ops_client2 = fs_util.get_mds_metrics(
-            client2, 0, fuse_mounting_dir_2, cephfs
+        pre_write_ops_client2 = get_mds_metrics_from_ranks(
+            [0, 1, 2], fs_util, client2, fuse_mounting_dir_2, cephfs
         )["counters"]
         log.info(f"pre_read_ops_client1: {pre_read_ops_client1}")
         log.info(f"pre_write_ops_client2: {pre_write_ops_client2}")
@@ -456,11 +563,11 @@ def run(ceph_cluster, **kw):
             "total_write_size",
         ]
         log.info("Get final MDS metrics after copying a file")
-        post_read_ops_client1 = fs_util.get_mds_metrics(
-            client1, 0, fuse_mounting_dir_1, cephfs
+        post_read_ops_client1 = get_mds_metrics_from_ranks(
+            [0, 1, 2], fs_util, client1, fuse_mounting_dir_1, cephfs
         )["counters"]
-        post_write_ops_client2 = fs_util.get_mds_metrics(
-            client2, 0, fuse_mounting_dir_2, cephfs
+        post_write_ops_client2 = get_mds_metrics_from_ranks(
+            [0, 1, 2], fs_util, client2, fuse_mounting_dir_2, cephfs
         )["counters"]
         log.info(f"post_read_ops_client1: {post_read_ops_client1}")
         log.info(f"post_write_ops_client2: {post_write_ops_client2}")

tests/cephfs/lib/cephfs_common_lib.py

@@ -45,13 +45,20 @@ def wait_for_healthy_ceph(self, client, wait_time):
         """
         ceph_healthy = 0
         end_time = datetime.datetime.now() + datetime.timedelta(seconds=wait_time)
+        accepted_list = [
+            "experiencing slow operations in BlueStore",
+            "Slow OSD heartbeats",
+        ]
         while ceph_healthy == 0 and (datetime.datetime.now() < end_time):
             if self.check_ceph_status(client, "HEALTH_OK"):
                 ceph_healthy = 1
             else:
                 out, _ = client.exec_command(sudo=True, cmd="ceph health detail")
-                if "experiencing slow operations in BlueStore" in str(out):
-                    log.info("Ignoring the known warning for Bluestore Slow ops")
+                if any(msg in str(out) for msg in accepted_list):
+                    log.info(
+                        "Ignoring the known warning for Bluestore Slow ops and OSD heartbeats"
+                    )
+                    log.warning("Cluster health can be OK, current state : %s", out)
                     ceph_healthy = 1
                 else:
                     log.info(