Create the resize osd test and the verification steps

Signed-off-by: Itzhak Kave <[email protected]>

ocs_ci/ocs/cluster.py

@@ -30,6 +30,7 @@
     ThinPoolUtilityWrong,
     TimeoutExpiredError,
     ResourceWrongStatusException,
+    CephHealthException,
 )
 from ocs_ci.ocs.resources import ocs, storage_cluster
 import ocs_ci.ocs.constants as constant
@@ -50,14 +51,24 @@
 from ocs_ci.framework import config
 from ocs_ci.ocs import ocp, constants, exceptions
 from ocs_ci.ocs.exceptions import PoolNotFound
-from ocs_ci.ocs.resources.pvc import get_all_pvc_objs
+from ocs_ci.ocs.resources.pvc import (
+    get_all_pvc_objs,
+    get_deviceset_pvcs,
+    get_deviceset_pvs,
+)
 from ocs_ci.ocs.ocp import OCP, wait_for_cluster_connectivity
 from ocs_ci.ocs.resources.ocs import OCS
 from ocs_ci.ocs.resources.pvc import PVC
 from ocs_ci.utility.connection import Connection
 from ocs_ci.utility.lvmo_utils import get_lvm_cluster_name
-from ocs_ci.ocs.resources.pod import get_mds_pods, wait_for_pods_to_be_running
+from ocs_ci.ocs.resources.pod import (
+    get_mds_pods,
+    wait_for_pods_to_be_running,
+    get_osd_pods,
+    delete_pods,
+)
 from ocs_ci.utility.decorators import switch_to_orig_index_at_last
+from ocs_ci.ocs.resources.pv import get_pv_size
 
 logger = logging.getLogger(__name__)
 
@@ -3268,3 +3279,279 @@ def client_clusters_health_check():
         client_cluster_health_check()
 
     logger.info("The client clusters health check passed successfully")
+
+
+def check_resources_state_post_resize_osd(old_osd_pods, old_osd_pvcs, old_osd_pvs):
+    """
+    Check that the pods, PVCs, and PVs are in the expected state post resizing the osd.
+    It will perform the following steps:
+        1. Check that the old osd pods are in a terminating state or deleted
+        2. Check that the new osd pods running, and we have exactly the same number
+           of osd pods as the old ones.
+        3. Check that the PVCs are in a Bound state
+        4. Check that the old PVC and PV names are equal to the current PVC and PV names
+
+    Args:
+        old_osd_pods (list): The old osd pod objects before resizing the osd
+        old_osd_pvcs (list): The old osd PVC objects before resizing the osd
+        old_osd_pvs (list): The old osd PV objects before resizing the osd
+
+    Raises:
+        ResourceWrongStatusException: If the pods, PVCs, and PVs are not in the expected state.
+
+    """
+    old_osd_pods_count = len(old_osd_pods)
+    logger.info("Wait for the OSD pods to reach the status Terminated or be deleted")
+    old_osd_pod_names = [p.name for p in old_osd_pods]
+    res = pod.wait_for_pods_to_be_in_statuses(
+        expected_statuses=[constants.STATUS_TERMINATING],
+        pod_names=old_osd_pod_names,
+        timeout=300,
+        sleep=20,
+    )
+    if not res:
+        raise ResourceWrongStatusException(
+            "The OSD pods failed to reach the status Terminated or be deleted"
+        )
+
+    logger.info("Check that the new OSD pods are running")
+    ocp_pod = OCP(kind=constants.POD, namespace=config.ENV_DATA["cluster_namespace"])
+    ocp_pod.wait_for_resource(
+        condition=constants.STATUS_RUNNING,
+        selector=constants.OSD_APP_LABEL,
+        resource_count=old_osd_pods_count,
+        timeout=300,
+        sleep=20,
+    )
+
+    logger.info(
+        f"Check that the number of the new OSD pods are exactly {old_osd_pods_count}"
+    )
+    for osd_pods in TimeoutSampler(timeout=180, sleep=10, func=get_osd_pods):
+        osd_pods_count = len(osd_pods)
+        logger.info(f"number of osd pods = {osd_pods_count}")
+        if old_osd_pods_count == osd_pods_count:
+            break
+
+    logger.info("Check that the PVCs are in a Bound state")
+    ocp_pvc = OCP(kind=constants.PVC, namespace=config.ENV_DATA["cluster_namespace"])
+    ocp_pvc.wait_for_resource(
+        timeout=30,
+        sleep=5,
+        condition=constants.STATUS_BOUND,
+        selector=constants.OSD_PVC_GENERIC_LABEL,
+        resource_count=len(old_osd_pvcs),
+    )
+
+    current_osd_pvcs = get_deviceset_pvcs()
+    old_pvc_names = [p.name for p in old_osd_pvcs]
+    current_pvc_names = [p.name for p in current_osd_pvcs]
+    logger.info(f"Old PVC names = {old_pvc_names}")
+    logger.info(f"Current PVC names = {current_pvc_names}")
+
+    current_osd_pvs = get_deviceset_pvs()
+    old_pv_names = [p.name for p in old_osd_pvs]
+    current_pv_names = [p.name for p in current_osd_pvs]
+    logger.info(f"Old PV names = {old_pv_names}")
+    logger.info(f"Current PV names = {current_pv_names}")
+
+    logger.info(
+        "Check that the old PVC and PV names are equal to the current PVC and PV names"
+    )
+    if not old_pvc_names == current_pvc_names:
+        raise ResourceWrongStatusException(
+            f"The old PVC names {old_pvc_names} are not equal to the "
+            f"current PVC names {current_pvc_names}"
+        )
+    if not old_pv_names == current_pv_names:
+        raise ResourceWrongStatusException(
+            f"The old PV names {old_pv_names} are not equal to the "
+            f"current PV names {current_pv_names}"
+        )
+
+
+def check_resources_size_post_resize_osd(expected_storage_size):
+    """
+    Check if the current storagecluster size, PVCs, and PVs are in the expected size.
+
+    Args:
+        expected_storage_size (str): The expected storage size after resizing the osd
+
+    Raises:
+        ResourceWrongStatusException: If the pods, PVCs, and PVs are not in the expected size
+
+    """
+    logger.info(f"The expected storage size is {expected_storage_size}")
+
+    current_storage_size = storage_cluster.get_storage_size()
+    logger.info(f"The current storage size is {current_storage_size}")
+    logger.info(
+        "Check that the current storage size equal to the expected storage size"
+    )
+    if storage_cluster.get_storage_size() != expected_storage_size:
+        raise ResourceWrongStatusException(
+            f"The current storage size {current_storage_size} is not equal "
+            f"to the expected size {expected_storage_size}"
+        )
+
+    logger.info(
+        "Check that the PVC and PV sizes are equal to the expected storage size"
+    )
+    current_osd_pvcs = get_deviceset_pvcs()
+    expected_storage_size_in_gb = convert_device_size(expected_storage_size, "GB")
+    pvc_sizes = [pvc.size for pvc in current_osd_pvcs]
+    logger.info(f"PVC sizes = {pvc_sizes}")
+    if not all([p_size == expected_storage_size_in_gb for p_size in pvc_sizes]):
+        raise ResourceWrongStatusException(
+            f"The PVC sizes are not equal to the expected storage size {expected_storage_size_in_gb}"
+        )
+
+    current_osd_pvs = get_deviceset_pvs()
+    pv_sizes = [get_pv_size(pv.get()) for pv in current_osd_pvs]
+    logger.info(f"PV sizes {pv_sizes}")
+    if not all([p_size == expected_storage_size_in_gb for p_size in pv_sizes]):
+        raise ResourceWrongStatusException(
+            f"The PV sizes are not equal to the expected storage size {expected_storage_size_in_gb}"
+        )
+
+
+def check_ceph_state_post_resize_osd(expected_storage_size):
+    """
+    Check the Ceph state post resize osd. It will perform the following steps:
+        1. Check that the Ceph capacity is equal to the expected storage size
+        2. Check the Ceph device classes and osd tree
+
+    Args:
+        expected_storage_size (str): The expected storage size after resizing the osd
+
+    Raises:
+        CephHealthException: If the Ceph capacity is not equal to the expected storage size, or
+            the Ceph device classes and osd tree checks failed
+
+    """
+    ceph_cluster = CephCluster()
+    ceph_capacity = ceph_cluster.get_ceph_capacity()
+    expected_storage_size_in_gb = convert_device_size(expected_storage_size, "GB")
+    logger.info(
+        f"Check that the Ceph capacity {ceph_capacity} is equal "
+        f"to the expected storage size {expected_storage_size_in_gb}"
+    )
+    if not int(ceph_capacity) == expected_storage_size_in_gb:
+        raise CephHealthException(
+            f"The Ceph capacity {ceph_capacity} is not equal to the "
+            f"expected storage size {expected_storage_size_in_gb}"
+        )
+
+    logger.info("Check the Ceph device classes and osd tree")
+    device_class = storage_cluster.get_device_class()
+    ct_pod = pod.get_ceph_tools_pod()
+    try:
+        storage_cluster.verify_storage_device_class(device_class)
+        storage_cluster.verify_device_class_in_osd_tree(ct_pod, device_class)
+    except AssertionError as ex:
+        raise CephHealthException(ex)
+    if not check_ceph_osd_tree():
+        raise CephHealthException("The OSD tree is not created/modified correctly")
+
+
+def base_ceph_verification_steps_post_resize_osd(
+    old_osd_pods, old_osd_pvcs, old_osd_pvs, expected_storage_size
+):
+    """
+    Check the Ceph verification steps post resize OSD.
+    It will perform the following steps:
+        1. Check the resources state post resize OSD
+        2. Check the resources size post resize OSD
+        3. Check the Ceph state post resize OSD
+
+    Args:
+        old_osd_pods (list): The old osd pod objects before resizing the osd
+        old_osd_pvcs (list): The old osd PVC objects before resizing the osd
+        old_osd_pvs (list): The old osd PV objects before resizing the osd
+        expected_storage_size (str): The expected storage size after resizing the osd
+
+    Raises:
+        ResourceWrongStatusException: If the resources are not in the expected state and size
+        CephHealthException: If Ceph is not in the expected state.
+
+    """
+    logger.info("Check the resources state post resize OSD")
+    check_resources_state_post_resize_osd(old_osd_pods, old_osd_pvcs, old_osd_pvs)
+    logger.info("Check the resources size post resize OSD")
+    check_resources_size_post_resize_osd(expected_storage_size)
+    logger.info("Check the Ceph state post resize OSD")
+    check_ceph_state_post_resize_osd(expected_storage_size)
+    logger.info("All the Ceph verification steps post resize osd finished successfully")
+
+
+def ceph_verification_steps_post_resize_osd(
+    old_osd_pods, old_osd_pvcs, old_osd_pvs, expected_storage_size
+):
+    """
+    Check the Ceph verification steps post resize OSD, as described in the above function:
+    'base_ceph_verification_steps_post_resize_osd'.
+    If we get one of the exceptions: 'ResourceWrongStatusException' and 'CephHealthException' it will
+    restart the osd pods and try again several times.
+    If it fails after the number of tries, it returns False. Otherwise, if it succeeded, it returns True.
+
+    Args:
+        old_osd_pods (list): The old osd pod objects before resizing the osd
+        old_osd_pvcs (list): The old osd PVC objects before resizing the osd
+        old_osd_pvs (list): The old osd PV objects before resizing the osd
+        expected_storage_size (str): The expected storage size after resizing the osd
+
+    Returns:
+        bool: True, if the Ceph verification steps post resize OSD succeeded after the given tries.
+            False, otherwise
+
+    """
+    num_of_tries = 5
+    for i in range(num_of_tries):
+        try:
+            base_ceph_verification_steps_post_resize_osd(
+                old_osd_pods, old_osd_pvcs, old_osd_pvs, expected_storage_size
+            )
+            return True
+        except (ResourceWrongStatusException, CephHealthException) as ex:
+            logger.warning(
+                f"The Ceph verification steps failed due to the error: {str(ex)}. "
+                f"Try to restart the OSD pods before the next iteration"
+            )
+            old_osd_pods = get_osd_pods()
+            delete_pods(old_osd_pods, wait=False)
+
+    logger.warning(
+        f"Failed to complete the Ceph verification steps post resize osd after {num_of_tries} tries"
+    )
+    return False
+
+
+def check_ceph_health_after_resize_osd(
+    ceph_health_tries=40, ceph_rebalance_timeout=900
+):
+    """
+    Check Ceph health after resize osd
+
+    Args:
+        ceph_health_tries (int): The number of tries to wait for the Ceph health to be OK.
+        ceph_rebalance_timeout (int): The time to wait for the Ceph cluster rebalanced.
+
+    """
+    if config.RUN.get("io_in_bg"):
+        logger.info(
+            "Increase the time to wait for Ceph health to be health OK, "
+            "because we run IO in the background"
+        )
+        additional_ceph_health_tries = int(config.RUN.get("io_load") * 1.3)
+        ceph_health_tries += additional_ceph_health_tries
+
+        additional_ceph_rebalance_timeout = int(config.RUN.get("io_load") * 100)
+        ceph_rebalance_timeout += additional_ceph_rebalance_timeout
+
+    ceph_health_check(
+        namespace=config.ENV_DATA["cluster_namespace"], tries=ceph_health_tries
+    )
+    ceph_cluster_obj = CephCluster()
+    assert ceph_cluster_obj.wait_for_rebalance(
+        timeout=ceph_rebalance_timeout
+    ), "Data re-balance failed to complete"

ocs_ci/ocs/resources/pod.py

@@ -3583,3 +3583,64 @@ def _check_if_pod_deleted(label, namespace):
         namespace=namespace,
     )
     sampler.wait_for_func_status(True)
+
+
+def calculate_md5sum_of_pod_files(pods_for_integrity_check, pod_file_name):
+    """
+    Calculate the md5sum of the pod files, and save it in the pod objects
+
+    Args:
+        pods_for_integrity_check (list): The list of the pod objects to calculate the md5sum
+        pod_file_name (str): The pod file name to save the md5sum
+
+    """
+    # Wait for IO to finish
+    logger.info("Wait for IO to finish on pods")
+    for pod_obj in pods_for_integrity_check:
+        pod_obj.get_fio_results()
+        logger.info(f"IO finished on pod {pod_obj.name}")
+        # Calculate md5sum
+        pod_file_name = (
+            pod_file_name
+            if (pod_obj.pvc.volume_mode == constants.VOLUME_MODE_FILESYSTEM)
+            else pod_obj.get_storage_path(storage_type="block")
+        )
+        logger.info(
+            f"Calculate the md5sum of the file {pod_file_name} in the pod {pod_obj.name}"
+        )
+        pod_obj.pvc.md5sum = cal_md5sum(
+            pod_obj,
+            pod_file_name,
+            pod_obj.pvc.volume_mode == constants.VOLUME_MODE_BLOCK,
+        )
+
+
+def verify_md5sum_on_pod_files(pods_for_integrity_check, pod_file_name):
+    """
+    Verify the md5sum of the pod files
+
+    Args:
+        pods_for_integrity_check (list): The list of the pod objects to verify the md5sum
+        pod_file_name (str): The pod file name to verify its md5sum
+
+    Raises:
+        AssertionError: If file doesn't exist or md5sum mismatch
+
+    """
+    for pod_obj in pods_for_integrity_check:
+        pod_file_name = (
+            pod_obj.get_storage_path(storage_type="block")
+            if (pod_obj.pvc.volume_mode == constants.VOLUME_MODE_BLOCK)
+            else pod_file_name
+        )
+        verify_data_integrity(
+            pod_obj,
+            pod_file_name,
+            pod_obj.pvc.md5sum,
+            pod_obj.pvc.volume_mode == constants.VOLUME_MODE_BLOCK,
+        )
+        logger.info(
+            f"Verified: md5sum of {pod_file_name} on pod {pod_obj.name} "
+            f"matches with the original md5sum"
+        )
+    logger.info("Data integrity check passed on all pods")