Rechercher

Ce contenu n'est pas disponible dans la langue sélectionnée.

Chapter 11. Management of NFS-Ganesha gateway using the Ceph Orchestrator (Limited Availability)

download PDF

As a storage administrator, you can use the Orchestrator with Cephadm in the backend to deploy the NFS-Ganesha gateway. Cephadm deploys NFS Ganesha using a predefined RADOS pool and optional namespace.

Note

This technology is Limited Availability. See the Deprecated functionality chapter for additional information.

Note

Red Hat supports CephFS exports only over the NFS v4.0+ protocol.

This section covers the following administrative tasks:

Prerequisites

  • A running Red Hat Ceph Storage cluster.
  • Root-level access to all the nodes.
  • Hosts are added to the cluster.
  • All manager, monitor and OSD daemons are deployed.

11.1. Creating the NFS-Ganesha cluster using the Ceph Orchestrator

You can create an NFS-Ganesha cluster using the mgr/nfs module of the Ceph Orchestrator. This module deploys the NFS cluster using Cephadm in the backend.

This creates a common recovery pool for all NFS-Ganesha daemons, new user based on clusterid, and a common NFS-Ganesha config RADOS object.

For each daemon, a new user and a common configuration is created in the pool. Although all the clusters have different namespaces with respect to the cluster names, they use the same recovery pool.

Prerequisites

  • A running Red Hat Ceph Storage cluster.
  • Hosts are added to the cluster.
  • All manager, monitor and OSD daemons are deployed.

Procedure

  1. Log into the Cephadm shell:

    Example

    [root@host01 ~]# cephadm shell

  2. Enable the mgr/nfs module:

    Example

    [ceph: root@host01 /]# ceph mgr module enable nfs

  3. Create the cluster:

    Syntax

    ceph nfs cluster create CLUSTER_NAME ["HOST_NAME_1 HOST_NAME_2 HOST_NAME_3"]

    The CLUSTER_NAME is an arbitrary string and HOST_NAME_1 is an optional string signifying the hosts to deploy NFS-Ganesha daemons.

    Example

    [ceph: root@host01 /]# ceph nfs cluster create nfsganesha "host01 host02"
    NFS Cluster Created Successful

    This creates an NFS_Ganesha cluster nfsganesha with one daemon on host01 and host02.

Verification

  • List the cluster details:

    Example

    [ceph: root@host01 /]# ceph nfs cluster ls

  • Show NFS-Ganesha cluster information:

    Syntax

    ceph nfs cluster info CLUSTER_NAME

    Example

    [ceph: root@host01 /]# ceph nfs cluster info nfsganesha

Additional Resources

11.2. Deploying the NFS-Ganesha gateway using the command line interface

You can use the Ceph Orchestrator with Cephadm in the backend to deploy the NFS-Ganesha gateway using the placement specification. In this case, you have to create a RADOS pool and create a namespace before deploying the gateway.

Note

Red Hat supports CephFS exports only over the NFS v4.0+ protocol.

Prerequisites

  • A running Red Hat Ceph Storage cluster.
  • Hosts are added to the cluster.
  • All manager, monitor and OSD daemons are deployed.

Procedure

  1. Log into the Cephadm shell:

    Example

    [root@host01 ~]# cephadm shell

  2. Create the RADOS pool namespace, and enable the application. For RBD pools, enable RBD.

    Syntax

    ceph osd pool create POOL_NAME
    ceph osd pool application enable POOL_NAME freeform/rgw/rbd/cephfs/nfs
    rbd pool init -p POOL_NAME

    Example

    [ceph: root@host01 /]# ceph osd pool create nfs-ganesha
    [ceph: root@host01 /]# ceph osd pool application enable nfs-ganesha nfs
    [ceph: root@host01 /]# rbd pool init -p nfs-ganesha

  3. Deploy NFS-Ganesha gateway using placement specification in the command line interface:

    Syntax

    ceph orch apply nfs SERVICE_ID --placement="NUMBER_OF_DAEMONS HOST_NAME_1 HOST_NAME_2 HOST_NAME_3"

    Example

    [ceph: root@host01 /]# ceph orch apply nfs foo --placement="2 host01 host02"

    This deploys an NFS-Ganesha cluster nfsganesha with one daemon on host01 and host02.

Verification

  • List the service:

    Example

    [ceph: root@host01 /]# ceph orch ls

  • List the hosts, daemons, and processes:

    Syntax

    ceph orch ps --daemon_type=DAEMON_NAME

    Example

    [ceph: root@host01 /]# ceph orch ps --daemon_type=nfs

Additional Resources

11.3. Deploying the NFS-Ganesha gateway using the service specification

You can use the Ceph Orchestrator with Cephadm in the backend to deploy the NFS-Ganesha gateway using the service specification. In this case, you have to create a RADOS pool and create a namespace before deploying the gateway.

Prerequisites

  • A running Red Hat Ceph Storage cluster.
  • Hosts are added to the cluster.

Procedure

  1. Create the nfs.yaml file:

    Example

    [root@host01 ~]# touch nfs.yaml

  2. Edit the nfs.yaml specification file.

    Syntax

    service_type: nfs
    service_id: SERVICE_ID
    placement:
      hosts:
        - HOST_NAME_1
        - HOST_NAME_2

    Example

    # cat nfs.yaml
    service_type: nfs
    service_id: foo
    placement:
      hosts:
        - host01
        - host02

    • Optional: Enable the NLM to support locking for NFS protocol v3 support by adding enable_nlm: true in the ganesha.yaml specification file.

      Syntax

      service_type: nfs
      service_id: SERVICE_ID
      placement:
        hosts:
          - HOST_NAME_1
          - HOST_NAME_2
      spec:
          enable_nlm: true

      Example

      # cat ganesha.yaml
      service_type: nfs
      service_id: foo
      placement:
        hosts:
          - host01
          - host02
      spec:
          enable_nlm: true

  3. Mount the YAML file under a directory in the container:

    Example

    [root@host01 ~]# cephadm shell --mount nfs.yaml:/var/lib/ceph/nfs.yaml

  4. Create the RADOS pool, namespace, and enable RBD:

    Syntax

    ceph osd pool create POOL_NAME
    ceph osd pool application enable POOL_NAME rbd
    rbd pool init -p POOL_NAME

    Example

    [ceph: root@host01 /]# ceph osd pool create nfs-ganesha
    [ceph: root@host01 /]# ceph osd pool application enable nfs-ganesha rbd
    [ceph: root@host01 /]# rbd pool init -p nfs-ganesha

  5. Navigate to the directory:

    Example

    [ceph: root@host01 /]# cd /var/lib/ceph/

  6. Deploy NFS-Ganesha gateway using service specification:

    Syntax

    ceph orch apply -i FILE_NAME.yaml

    Example

    [ceph: root@host01 ceph]# ceph orch apply -i nfs.yaml

Verification

  • List the service:

    Example

    [ceph: root@host01 /]# ceph orch ls

  • List the hosts, daemons, and processes:

    Syntax

    ceph orch ps --daemon_type=DAEMON_NAME

    Example

    [ceph: root@host01 /]# ceph orch ps --daemon_type=nfs

Additional Resources

11.4. Implementing HA for CephFS/NFS service (Technology Preview)

You can deploy NFS with a high-availability (HA) front-end, virtual IP, and load balancer, by using the --ingress flag and by specifying a virtual IP address. This deploys a combination of keepalived and haproxy and provides a high-availability NFS frontend for the NFS service.

When a cluster is created with --ingress flag, an ingress service is additionally deployed to provide load balancing and high-availability for the NFS servers. A virtual IP is used to provide a known, stable NFS endpoint that all NFS clients can use to mount. Ceph handles the details of redirecting NFS traffic on the virtual IP to the appropriate backend NFS servers and redeploys NFS servers when they fail.

Deploying an ingress service for an existing service provides:

  • A stable, virtual IP that can be used to access the NFS server.
  • Load distribution across multiple NFS gateways.
  • Failover between hosts in the event of a host failure.
Important

HA for CephFS/NFS is a Technology Preview feature only. Technology Preview features are not supported with Red Hat production service level agreements (SLAs), might not be functionally complete, and Red Hat does not recommend using them for production. These features provide early access to upcoming product features, enabling customers to test functionality and provide feedback during the development process. See the support scope for Red Hat Technology Preview features for more details.

Important

When an ingress service is deployed in front of the NFS cluster, the backend NFS-ganesha servers will see the haproxy’s IP address and not the client’s IP address. As a result, if you are restricting client access based on IP address, access restrictions for NFS exports will not work as expected.

Note

If the active NFS server serving a client goes down, the client’s I/Os are interrupted until the replacement for the active NFS server is online and the NFS cluster is active again.

Prerequisites

  • A running Red Hat Ceph Storage cluster.
  • Hosts are added to the cluster.
  • All the manager, monitor, and OSD daemons are deployed.
  • Ensure the NFS module is enabled.

Procedure

  1. Log into the Cephadm shell:

    Example

    [root@host01 ~]# cephadm shell

  2. Create the NFS cluster with the --ingress flag:

    Syntax

    ceph nfs cluster create CLUSTER_ID [PLACEMENT] [--port PORT_NUMBER] [--ingress --virtual-ip IP_ADDRESS/CIDR_PREFIX]

    • Replace CLUSTER_ID with a unique string to name the NFS Ganesha cluster.
    • Replace PLACEMENT with the number of NFS servers to deploy and the host or hosts that you want to deploy the NFS Ganesha daemon containers on.
    • Use the --port PORT_NUMBER flag to deploy NFS on a port other than the default port of 12049.

      Note

      With ingress mode, the high-availability proxy takes port 2049 and NFS services are deployed on 12049.

    • The --ingress flag combined with the --virtual-ip flag, deploys NFS with a high-availability front-end (virtual IP and load balancer).
    • Replace --virtual-ip IP_ADDRESS with an IP address to provide a known, stable NFS endpoint that all clients can use to mount NFS exports. The --virtual-ip must include a CIDR prefix length. The virtual IP will normally be configured on the first identified network interface that has an existing IP in the same subnet.

      Note

      The number of hosts you allocate for the NFS service must be greater than the number of active NFS servers you request to deploy, specified by the placement: count parameter. In the below example, one active NFS server is requested and two hosts are allocated.

      Example

      [ceph: root@host01 /]# ceph nfs cluster create mycephnfs "1 host02 host03" --ingress --virtual-ip 10.10.128.75/22

      Note

      Deployment of NFS daemons and the ingress service is asynchronous and the command might return before the services have completely started.

  3. Check that the services have successfully started:

    Syntax

    ceph orch ls --service_name=nfs.CLUSTER_ID
    ceph orch ls --service_name=ingress.nfs.CLUSTER_ID

    Example

    [ceph: root@host01 /]# ceph orch ls --service_name=nfs.mycephnfs
    
    NAME           PORTS    RUNNING  REFRESHED  AGE  PLACEMENT
    nfs.mycephnfs  ?:12049      2/2  0s ago     20s  host02;host03
    
    [ceph: root@host01 /]# ceph orch ls --service_name=ingress.nfs.mycephnfs
    
    NAME                   PORTS                  RUNNING  REFRESHED  AGE  PLACEMENT
    ingress.nfs.mycephnfs  10.10.128.75:2049,9049      4/4   46s ago    73s  count:2

Verification

  • View the IP endpoints, IPs for the individual NFS daemons, and the virtual IP for the ingress service:

    Syntax

    ceph nfs cluster info CLUSTER_NAME

    Example

    [ceph: root@host01 /]# ceph nfs cluster info mycephnfs
    
    {
        "mycephnfs": {
            "virtual_ip": "10.10.128.75",
            "backend": [
                {
                    "hostname": "host02",
                    "ip": "10.10.128.69",
                    "port": 12049
                },
                {
                    "hostname": "host03",
                    "ip": "10.10.128.70",
                    "port": 12049
                }
            ],
            "port": 2049,
            "monitor_port": 9049
        }
    }

  • List the hosts and processes:

    Example

    [ceph: root@host01 /]# ceph orch ps | grep nfs
    
    haproxy.nfs.cephnfs.host01.rftylv     host01  *:2049,9000  running (11m)    10m ago  11m    23.2M        -  2.2.19-7ea3822   5e6a41d77b38  f8cc61dc827e
    haproxy.nfs.cephnfs.host02.zhtded     host02  *:2049,9000  running (11m)    53s ago  11m    21.3M        -  2.2.19-7ea3822   5e6a41d77b38  4cad324e0e23
    keepalived.nfs.cephnfs.host01.zktmsk  host01               running (11m)    10m ago  11m    2349k        -  2.1.5            18fa163ab18f  66bf39784993
    keepalived.nfs.cephnfs.host02.vyycvp  host02               running (11m)    53s ago  11m    2349k        -  2.1.5            18fa163ab18f  1ecc95a568b4
    nfs.cephnfs.0.0.host02.fescmw         host02  *:12049      running (14m)     3m ago  14m    76.9M        -  3.5              cef6e7959b0a  bb0e4ee9484e
    nfs.cephnfs.1.0.host03.avaddf         host03  *:12049      running (14m)     3m ago  14m    74.3M        -  3.5              cef6e7959b0a  ea02c0c50749

Additional resources

11.5. Upgrading a standalone CephFS/NFS cluster for HA

As a storage administrator, you can upgrade a standalone storage cluster to a high-availability (HA) cluster by deploying the ingress service on an existing NFS service.

Prerequisites

  • A running Red Hat Ceph Storage cluster with an existing NFS service.
  • Hosts are added to the cluster.
  • All the manager, monitor, and OSD daemons are deployed.
  • Ensure the NFS module is enabled.

Procedure

  1. Log into the Cephadm shell:

    Example

    [root@host01 ~]# cephadm shell

  2. List existing NFS clusters:

    Example

    [ceph: root@host01 /]# ceph nfs cluster ls
    
    mynfs

    Note

    If a standalone NFS cluster is created on one node, you need to increase it to two or more nodes for HA. To increase the NFS service, edit the nfs.yaml file and increase the placements with the same port number.

    The number of hosts you allocate for the NFS service must be greater than the number of active NFS servers you request to deploy, specified by the placement: count parameter.

    Syntax

    service_type: nfs
    service_id: SERVICE_ID
    placement:
      hosts:
        - HOST_NAME_1
        - HOST_NAME_2
      count: COUNT
    spec:
      port: PORT_NUMBER

    Example

    service_type: nfs
    service_id: mynfs
    placement:
      hosts:
        - host02
        - host03
      count: 1
    spec:
      port: 12345

    In this example, the existing NFS service is running on port 12345 and an additional node is added to the NFS cluster with the same port.

    Apply the nfs.yaml service specification changes to upgrade to a two node NFS service:

    Example

    [ceph: root@host01 ceph]# ceph orch apply -i nfs.yaml

  3. Edit the ingress.yaml specification file with the existing NFS cluster ID:

    Syntax

    service_type: SERVICE_TYPE
    service_id: SERVICE_ID
    placement:
      count: PLACEMENT
    spec:
      backend_service: SERVICE_ID_BACKEND 1
      frontend_port: FRONTEND_PORT
      monitor_port: MONITOR_PORT 2
      virtual_ip: VIRTUAL_IP_WITH_CIDR

    1
    SERVICE_ID_BACKEND should include a PORT property that is not 2049 to avoid conflicting with the ingress service, which could be placed on the same host.
    2
    MONITOR_PORT is used to access the haproxy load status page.

    Example

    service_type: ingress
    service_id: nfs.mynfs
    placement:
      count: 2
    spec:
      backend_service: nfs.mynfs
      frontend_port: 2049
      monitor_port: 9000
      virtual_ip: 10.10.128.75/22

  4. Deploy the ingress service:

    Example

    [ceph: root@host01 /]# ceph orch apply -i ingress.yaml

    Note

    Deployment of NFS daemons and the ingress service is asynchronous and the command might return before the services have completely started.

  5. Check that the ingress services have successfully started:

    Syntax

    ceph orch ls --service_name=ingress.nfs.CLUSTER_ID

    Example

    [ceph: root@host01 /]# ceph orch ls --service_name=ingress.nfs.mynfs
    
    NAME                   PORTS                  RUNNING  REFRESHED  AGE  PLACEMENT
    ingress.nfs.mynfs  10.10.128.75:2049,9000      4/4   4m ago     22m  count:2

Verification

  • View the IP endpoints, IPs for the individual NFS daemons, and the virtual IP for the ingress service:

    Syntax

    ceph nfs cluster info CLUSTER_NAME

    Example

    [ceph: root@host01 /]# ceph nfs cluster info mynfs
    
    {
        "mynfs": {
            "virtual_ip": "10.10.128.75",
            "backend": [
                {
                    "hostname": "host02",
                    "ip": "10.10.128.69",
                    "port": 12049
                },
                {
                    "hostname": "host03",
                    "ip": "10.10.128.70",
                    "port": 12049
                }
            ],
            "port": 2049,
            "monitor_port": 9049
        }
    }

  • List the hosts and processes:

    Example

    [ceph: root@host01 /]# ceph orch ps | grep nfs
    
    haproxy.nfs.mynfs.host01.ruyyhq     host01  *:2049,9000  running (27m)     6m ago  34m    9.85M        -  2.2.19-7ea3822   5e6a41d77b38  328d27b3f706
    haproxy.nfs.mynfs.host02.ctrhha     host02  *:2049,9000  running (34m)     6m ago  34m    4944k        -  2.2.19-7ea3822   5e6a41d77b38  4f4440dbfde9
    keepalived.nfs.mynfs.host01.fqgjxd  host01               running (27m)     6m ago  34m    31.2M        -  2.1.5            18fa163ab18f  0e22b2b101df
    keepalived.nfs.mynfs.host02.fqzkxb  host02               running (34m)     6m ago  34m    17.5M        -  2.1.5            18fa163ab18f  c1e3cc074cf8
    nfs.mynfs.0.0.host02.emoaut         host02  *:12345      running (37m)     6m ago  37m    82.7M        -  3.5              91322de4f795  2d00faaa2ae5
    nfs.mynfs.1.0.host03.nsxcfd         host03  *:12345      running (37m)     6m ago  37m    81.1M        -  3.5              91322de4f795  d4bda4074f17

Additional resources

11.6. Deploying HA for CephFS/NFS using a specification file

You can deploy HA for CephFS/NFS using a specification file by first deploying an NFS service and then deploying ingress to the same NFS service.

Prerequisites

  • A running Red Hat Ceph Storage cluster.
  • Hosts are added to the cluster.
  • All the manager, monitor, and OSD daemons are deployed.
  • Ensure the NFS module is enabled.

Procedure

  1. Log into the Cephadm shell:

    Example

    [root@host01 ~]# cephadm shell

  2. Ensure the NFS module is enabled:

    Example

    [ceph: root@host01 /]# ceph mgr module ls | more

  3. Exit out of the Cephadm shell and create the nfs.yaml file:

    Example

    [root@host01 ~]# touch nfs.yaml

  4. Edit the nfs.yaml file to include the following details:

    Syntax

    service_type: nfs
    service_id: SERVICE_ID
    placement:
      hosts:
        - HOST_NAME_1
        - HOST_NAME_2
      count: COUNT
    spec:
      port: PORT_NUMBER

    Note

    The number of hosts you allocate for the NFS service must be greater than the number of active NFS servers you request to deploy, specified by the placement: count parameter.

    Example

    service_type: nfs
    service_id: cephfsnfs
    placement:
      hosts:
        - host02
        - host03
      count: 1
    spec:
      port: 12345

    In this example, the server is run on the non-default port of 12345, instead of the default port of 2049, on host02 and host03.

  5. Mount the YAML file under a directory in the container:

    Example

    [root@host01 ~]# cephadm shell --mount nfs.yaml:/var/lib/ceph/nfs.yaml

  6. Log into the Cephadm shell and navigate to the directory:

    Example

    [ceph: root@host01 /]# cd /var/lib/ceph/

  7. Deploy the NFS service using service specification:

    Syntax

    ceph orch apply -i FILE_NAME.yaml

    Example

    [ceph: root@host01 ceph]# ceph orch apply -i nfs.yaml

    Note

    Deployment of the NFS service is asynchronous and the command might return before the services have completely started.

  8. Check that the NFS services have successfully started:

    Syntax

    ceph orch ls --service_name=nfs.CLUSTER_ID

    Example

    [ceph: root@host01 /]# ceph orch ls --service_name=nfs.cephfsnfs
    
    NAME           PORTS    RUNNING  REFRESHED  AGE  PLACEMENT
    nfs.cephfsnfs  ?:12345      2/2  3m ago     13m  host02;host03

  9. Exit out of the Cephadm shell and create the ingress.yaml file:

    Example

    [root@host01 ~]# touch ingress.yaml

  10. Edit the ingress.yaml file to include the following details:

    Syntax

    service_type: SERVICE_TYPE
    service_id: SERVICE_ID
    placement:
      count: PLACEMENT
    spec:
      backend_service: SERVICE_ID_BACKEND
      frontend_port: FRONTEND_PORT
      monitor_port: MONITOR_PORT
      virtual_ip: VIRTUAL_IP_WITH_CIDR

    Example

    service_type: ingress
    service_id: nfs.cephfsnfs
    placement:
      count: 2
    spec:
      backend_service: nfs.cephfsnfs
      frontend_port: 2049
      monitor_port: 9000
      virtual_ip: 10.10.128.75/22

    Note

    In this example, placement: count: 2 deploys the keepalived and haproxy service on random nodes. To specify the nodes to deploy keepalived and haproxy on, use the placement: hosts option:

    Example

    service_type: ingress
    service_id: nfs.cephfsnfs
    placement:
      hosts:
        - host02
        - host03

  11. Mount the YAML file under a directory in the container:

    Example

    [root@host01 ~]# cephadm shell --mount ingress.yaml:/var/lib/ceph/ingress.yaml

  12. Log into the Cephadm shell and navigate to the directory:

    Example

    [ceph: root@host01 /]# cd /var/lib/ceph/

  13. Deploy the ingress service using service specification:

    Syntax

    ceph orch apply -i FILE_NAME.yaml

    Example

    [ceph: root@host01 ceph]# ceph orch apply -i ingress.yaml

  14. Check that the ingress services have successfully started:

    Syntax

    ceph orch ls --service_name=ingress.nfs.CLUSTER_ID

    Example

    [ceph: root@host01 /]# ceph orch ls --service_name=ingress.nfs.cephfsnfs
    
    NAME                   PORTS                  RUNNING  REFRESHED  AGE  PLACEMENT
    ingress.nfs.cephfsnfs  10.10.128.75:2049,9000      4/4   4m ago     22m  count:2

Verification

  • View the IP endpoints, IPs for the individual NFS daemons, and the virtual IP for the ingress service:

    Syntax

    ceph nfs cluster info CLUSTER_NAME

    Example

    [ceph: root@host01 /]# ceph nfs cluster info cephfsnfs
    
    {
        "cephfsnfs": {
            "virtual_ip": "10.10.128.75",
            "backend": [
                {
                    "hostname": "host02",
                    "ip": "10.10.128.69",
                    "port": 12345
                },
                {
                    "hostname": "host03",
                    "ip": "10.10.128.70",
                    "port": 12345
                }
            ],
            "port": 2049,
            "monitor_port": 9049
        }
    }

  • List the hosts and processes:

    Example

    [ceph: root@host01 /]# ceph orch ps | grep nfs
    
    haproxy.nfs.cephfsnfs.host01.ruyyhq     host01  *:2049,9000  running (27m)     6m ago  34m    9.85M        -  2.2.19-7ea3822   5e6a41d77b38  328d27b3f706
    haproxy.nfs.cephfsnfs.host02.ctrhha     host02  *:2049,9000  running (34m)     6m ago  34m    4944k        -  2.2.19-7ea3822   5e6a41d77b38  4f4440dbfde9
    keepalived.nfs.cephfsnfs.host01.fqgjxd  host01               running (27m)     6m ago  34m    31.2M        -  2.1.5            18fa163ab18f  0e22b2b101df
    keepalived.nfs.cephfsnfs.host02.fqzkxb  host02               running (34m)     6m ago  34m    17.5M        -  2.1.5            18fa163ab18f  c1e3cc074cf8
    nfs.cephfsnfs.0.0.host02.emoaut         host02  *:12345      running (37m)     6m ago  37m    82.7M        -  3.5              91322de4f795  2d00faaa2ae5
    nfs.cephfsnfs.1.0.host03.nsxcfd         host03  *:12345      running (37m)     6m ago  37m    81.1M        -  3.5              91322de4f795  d4bda4074f17

Additional resources

11.7. Updating the NFS-Ganesha cluster using the Ceph Orchestrator

You can update the NFS-Ganesha cluster by changing the placement of the daemons on the hosts using the Ceph Orchestrator with Cephadm in the backend.

Prerequisites

  • A running Red Hat Ceph Storage cluster.
  • Hosts are added to the cluster.
  • All manager, monitor and OSD daemons are deployed.
  • NFS-Ganesha cluster created using the mgr/nfs module.

Procedure

  1. Log into the Cephadm shell:

    Example

    [root@host01 ~]# cephadm shell

  2. Update the cluster:

    Syntax

    ceph orch apply nfs CLUSTER_NAME ["HOST_NAME_1,HOST_NAME_2,HOST_NAME_3"]

    The CLUSTER_NAME is an arbitrary string, HOST_NAME_1 is an optional string signifying the hosts to update the deployed NFS-Ganesha daemons.

    Example

    [ceph: root@host01 /]# ceph orch apply nfs nfsganesha "host02"

    This updates the nfsganesha cluster on host02.

Verification

  • List the cluster details:

    Example

    [ceph: root@host01 /]# ceph nfs cluster ls

  • Show NFS-Ganesha cluster information:

    Syntax

    ceph nfs cluster info CLUSTER_NAME

    Example

    [ceph: root@host01 /]# ceph nfs cluster info nfsganesha

  • List the hosts, daemons, and processes:

    Syntax

    ceph orch ps --daemon_type=DAEMON_NAME

    Example

    [ceph: root@host01 /]# ceph orch ps --daemon_type=nfs

Additional Resources

11.8. Viewing the NFS-Ganesha cluster information using the Ceph Orchestrator

You can view the information of the NFS-Ganesha cluster using the Ceph Orchestrator. You can get the information about all the NFS Ganesha clusters or specific clusters with their port, IP address and the name of the hosts on which the cluster is created.

Prerequisites

  • A running Red Hat Ceph Storage cluster.
  • Hosts are added to the cluster.
  • All manager, monitor and OSD daemons are deployed.
  • NFS-Ganesha cluster created using the mgr/nfs module.

Procedure

  1. Log into the Cephadm shell:

    Example

    [root@host01 ~]# cephadm shell

  2. View the NFS-Ganesha cluster information:

    Syntax

    ceph nfs cluster info CLUSTER_NAME

    Example

    [ceph: root@host01 /]# ceph nfs cluster info nfsganesha
    
    {
        "nfsganesha": [
            {
                "hostname": "host02",
                "ip": [
                    "10.10.128.70"
                ],
                "port": 2049
            }
        ]
    }

Additional Resources

11.9. Fetching the NFS-Ganesha cluster logs using the Ceph Orchestrator

With the Ceph Orchestrator, you can fetch the NFS-Ganesha cluster logs. You need to be on the node where the service is deployed.

Prerequisites

  • A running Red Hat Ceph Storage cluster.
  • Cephadm installed on the nodes where NFS is deployed.
  • Root-level access to all the nodes.
  • Hosts are added to the cluster.
  • NFS-Ganesha cluster created using the mgr/nfs module.

Procedure

  1. As a root user, fetch the FSID of the storage cluster:

    Example

    [root@host03 ~]# cephadm ls

    Copy the FSID and the name of the service.

  2. Fetch the logs:

    Syntax

    cephadm logs --fsid FSID --name SERVICE_NAME

    Example

    [root@host03 ~]# cephadm logs --fsid 499829b4-832f-11eb-8d6d-001a4a000635 --name nfs.foo.host03

Additional Resources

11.10. Setting custom NFS-Ganesha configuration using the Ceph Orchestrator

The NFS-Ganesha cluster is defined in default configuration blocks. Using Ceph Orchestrator you can customize the configuration and that will have precedence over the default configuration blocks.

Prerequisites

  • A running Red Hat Ceph Storage cluster.
  • Hosts are added to the cluster.
  • All manager, monitor and OSD daemons are deployed.
  • NFS-Ganesha cluster created using the mgr/nfs module.

Procedure

  1. Log into the Cephadm shell:

    Example

    [root@host01 ~]# cephadm shell

  2. The following is an example of the default configuration of NFS-Ganesha cluster:

    Example

    # {{ cephadm_managed }}
    NFS_CORE_PARAM {
            Enable_NLM = false;
            Enable_RQUOTA = false;
            Protocols = 4;
    }
    
    MDCACHE {
            Dir_Chunk = 0;
    }
    
    EXPORT_DEFAULTS {
            Attr_Expiration_Time = 0;
    }
    
    NFSv4 {
            Delegations = false;
            RecoveryBackend = 'rados_cluster';
            Minor_Versions = 1, 2;
    }
    
    RADOS_KV {
            UserId = "{{ user }}";
            nodeid = "{{ nodeid}}";
            pool = "{{ pool }}";
            namespace = "{{ namespace }}";
    }
    
    RADOS_URLS {
            UserId = "{{ user }}";
            watch_url = "{{ url }}";
    }
    
    RGW {
            cluster = "ceph";
            name = "client.{{ rgw_user }}";
    }
    
    %url    {{ url }}

  3. Customize the NFS-Ganesha cluster configuration. The following are two examples for customizing the configuration:

    • Change the log level:

      Example

      LOG {
       COMPONENTS {
           ALL = FULL_DEBUG;
       }
      }

    • Add custom export block:

      1. Create the user.

        Note

        User specified in FSAL blocks should have proper caps for NFS-Ganesha daemons to access the Ceph cluster.

        Syntax

        ceph auth get-or-create client.USER_ID mon 'allow r' osd 'allow rw pool=.nfs namespace=NFS_CLUSTER_NAME, allow rw tag cephfs data=FS_NAME' mds 'allow rw path=EXPORT_PATH'

        Example

        [ceph: root@host01 /]# ceph auth get-or-create client.f64f341c-655d-11eb-8778-fa163e914bcc mon 'allow r' osd 'allow rw pool=.nfs namespace=nfs_cluster_name, allow rw tag cephfs data=fs_name' mds 'allow rw path=export_path'

      2. Navigate to the following directory:

        Syntax

        cd /var/lib/ceph/DAEMON_PATH/

        Example

        [ceph: root@host01 /]# cd /var/lib/ceph/nfs/

        If the nfs directory does not exist, create a directory in the path.

      3. Create a new configuration file:

        Syntax

        touch PATH_TO_CONFIG_FILE

        Example

        [ceph: root@host01 nfs]#  touch nfs-ganesha.conf

      4. Edit the configuration file by adding the custom export block. It creates a single export and that is managed by the Ceph NFS export interface.

        Syntax

        EXPORT {
          Export_Id = NUMERICAL_ID;
          Transports = TCP;
          Path = PATH_WITHIN_CEPHFS;
          Pseudo = BINDING;
          Protocols = 4;
          Access_Type = PERMISSIONS;
          Attr_Expiration_Time = 0;
          Squash = None;
          FSAL {
            Name = CEPH;
            Filesystem = "FILE_SYSTEM_NAME";
            User_Id = "USER_NAME";
            Secret_Access_Key = "USER_SECRET_KEY";
          }
        }

        Example

        EXPORT {
          Export_Id = 100;
          Transports = TCP;
          Path = /;
          Pseudo = /ceph/;
          Protocols = 4;
          Access_Type = RW;
          Attr_Expiration_Time = 0;
          Squash = None;
          FSAL {
            Name = CEPH;
            Filesystem = "filesystem name";
            User_Id = "user id";
            Secret_Access_Key = "secret key";
          }
        }

  4. Apply the new configuration the cluster:

    Syntax

    ceph nfs cluster config set CLUSTER_NAME -i PATH_TO_CONFIG_FILE

    Example

    [ceph: root@host01 nfs]# ceph nfs cluster config set nfs-ganesha -i /var/lib/ceph/nfs/nfs-ganesha.conf

    This also restarts the service for the custom configuration.

Verification

  • List the service:

    Example

    [ceph: root@host01 /]# ceph orch ls

  • List the hosts, daemons, and processes:

    Syntax

    ceph orch ps --daemon_type=DAEMON_NAME

    Example

    [ceph: root@host01 /]# ceph orch ps --daemon_type=nfs

  • Verify the custom configuration:

    Syntax

    /bin/rados -p POOL_NAME -N CLUSTER_NAME get userconf-nfs.CLUSTER_NAME -

    Example

    [ceph: root@host01 /]# /bin/rados -p nfs-ganesha -N nfsganesha get userconf-nfs.nfsganesha -

Additional Resources

11.11. Resetting custom NFS-Ganesha configuration using the Ceph Orchestrator

Using the Ceph Orchestrator, you can reset the user-defined configuration to the default configuration.

Prerequisites

  • A running Red Hat Ceph Storage cluster.
  • Hosts are added to the cluster.
  • All manager, monitor and OSD daemons are deployed.
  • NFS-Ganesha deployed using the mgr/nfs module.
  • Custom NFS cluster configuration is set-up

Procedure

  1. Log into the Cephadm shell:

    Example

    [root@host01 ~]# cephadm shell

  2. Reset the NFS-Ganesha configuration:

    Syntax

    ceph nfs cluster config reset CLUSTER_NAME

    Example

    [ceph: root@host01 /]# ceph nfs cluster config reset nfs-cephfs

Verification

  • List the service:

    Example

    [ceph: root@host01 /]# ceph orch ls

  • List the hosts, daemons, and processes:

    Syntax

    ceph orch ps --daemon_type=DAEMON_NAME

    Example

    [ceph: root@host01 /]# ceph orch ps --daemon_type=nfs

  • Verify the custom configuration is deleted:

    Syntax

    /bin/rados -p POOL_NAME -N CLUSTER_NAME get userconf-nfs.CLUSTER_NAME -

    Example

    [ceph: root@host01 /]# /bin/rados -p nfs-ganesha -N nfsganesha get userconf-nfs.nfsganesha -

Additional Resources

11.12. Deleting the NFS-Ganesha cluster using the Ceph Orchestrator

You can use the Ceph Orchestrator with Cephadm in the backend to delete the NFS-Ganesha cluster.

Prerequisites

  • A running Red Hat Ceph Storage cluster.
  • Hosts are added to the cluster.
  • All manager, monitor and OSD daemons are deployed.
  • NFS-Ganesha cluster created using the mgr/nfs module.

Procedure

  1. Log into the Cephadm shell:

    Example

    [root@host01 ~]# cephadm shell

  2. Delete the cluster:

    Syntax

    ceph nfs cluster rm CLUSTER_NAME

    The CLUSTER_NAME is an arbitrary string.

    Example

    [ceph: root@host01 /]# ceph nfs cluster rm nfsganesha
    NFS Cluster Deleted Successfully

    Note

    The delete option is deprecated and you need to use rm to delete an NFS cluster.

Verification

  • List the cluster details:

    Example

    [ceph: root@host01 /]# ceph nfs cluster ls

Additional Resources

11.13. Removing the NFS-Ganesha gateway using the Ceph Orchestrator

You can remove the NFS-Ganesha gateway using the ceph orch rm command.

Prerequisites

  • A running Red Hat Ceph Storage cluster.
  • Root-level access to all the nodes.
  • Hosts are added to the cluster.
  • At least one NFS-Ganesha gateway deployed on the hosts.

Procedure

  1. Log into the Cephadm shell:

    Example

    [root@host01 ~]# cephadm shell

  2. List the service:

    Example

    [ceph: root@host01 /]# ceph orch ls

  3. Remove the service:

    Syntax

    ceph orch rm SERVICE_NAME

    Example

    [ceph: root@host01 /]# ceph orch rm nfs.foo

Verification

  • List the hosts, daemons, and processes:

    Syntax

    ceph orch ps

    Example

    [ceph: root@host01 /]# ceph orch ps

Additional Resources

11.14. Kerberos integration

Kerberos is a computer network security protocol that provides a centralized authentication server which authenticates users to servers and vice versa across an untrusted network. In Kerberos Authentication, the server and database is used for client authentication.

11.14.1. Setting up the KDC (as per requirement)

Kerberos runs as a third-party trusted server known as the Key Distribution Center (KDC) in which each user and service on the network is a principal. The KDC holds information about all its clients (user principals, service principals); hence, it needs to be secure. In a Kerberos setup, as the KDC is a single point of failure, it is recommended to have one master KDC and multiple slave KDCs.

Prerequisites

Verify if the following changes are done in the /etc/hosts file. Add domain names if required.

[root@chost ~]# cat /etc/hosts

127.0.0.1   localhost localhost.localdomain localhost4 localhost4.localdomain4
::1         localhost localhost.localdomain localhost6 localhost6.localdomain6
10.0.208.97	        ceph-node1-installer.ibm.com	ceph-node1-installer
10.0.210.243	ceph-node2.ibm.com	ceph-node2
10.0.208.63	        ceph-node3.ibm.com	ceph-node3
10.0.210.222	ceph-node4.ibm.com	ceph-node4
10.0.210.235	ceph-node5.ibm.com	ceph-node5
10.0.209.87	        ceph-node6.ibm.com	ceph-node6
10.0.208.89	        ceph-node7.ibm.com	ceph-node7
Important

Ensure that domain name is present for all the involved nodes in the setup (all nodes in Ceph cluster and all NFS client nodes).

Procedure

Follow the below steps to install and configure KDC. Skip this part if you already have KDC installed and configured.

  1. Check if the required RPMs are installed on the machine where you want to setup the KDC.

    [root@host ~]# rpm -qa | grep krb5
    
    krb5-libs-1.20.1-9.el9_2.x86_64
    krb5-pkinit-1.20.1-9.el9_2.x86_64
    krb5-server-1.20.1-9.el9_2.x86_64
    krb5-server-ldap-1.20.1-9.el9_2.x86_64
    krb5-devel-1.20.1-9.el9_2.x86_64
    krb5-workstation-1.20.1-9.el9_2.x86_64
    Note
    • It is better to have domain name in accordance with Kerberos REALM name. For example, Realm - PUNE.IBM.COM, Admin principal - admin/admin.
    • Edit the installed configuration files to reflect the new KDC. Note that the KDC can be provided as either IP address or DNS name.
  2. Update the krb5.conf file:

    Note

    You need to update all the realms (default_realm and domain_realm) with the kdc and admin_server IP in the krb5.conf file.

    [root@host ~]# cat /etc/krb5.conf
    
    # To opt out of the system crypto-policies configuration of krb5, remove the
    # symlink at /etc/krb5.conf.d/crypto-policies which will not be recreated.
    
    includedir /etc/krb5.conf.d/
    [logging]
        default = [FILE:/var/log/krb5libs.log](file:///var/log/krb5libs.log)
        kdc = [FILE:/var/log/krb5kdc.log](file:///var/log/krb5kdc.log)
        admin_server = [FILE:/var/log/kadmind.log](file:///var/log/kadmind.log)
    [libdefaults]
        dns_lookup_realm = false
        ticket_lifetime = 24h
        renew_lifetime = 7d
        forwardable = true
        rdns = false
        pkinit_anchors = [FILE:/etc/pki/tls/certs/ca-bundle.crt](file:///etc/pki/tls/certs/ca-bundle.crt)
        spake_preauth_groups = edwards25519
        dns_canonicalize_hostname = fallback
        qualify_shortname = ""
        default_realm = PUNE.IBM.COM
        default_ccache_name = KEYRING:persistent:%{uid}
    [realms]
        PUNE.IBM.COM = {
           kdc = 10.0.210.222:88
           admin_server = 10.0.210.222:749
        }
    [domain_realm]
      .redhat.com = PUNE.IBM.COM
      redhat.com = PUNE.IBM.COM
  3. Update the krb5.conf file:

    Note

    You need to update the realms in the kdc.conf file.

    [root@host ~]# cat /var/kerberos/krb5kdc/kdc.conf
    
     [kdcdefaults]
     kdc_ports = 88
     kdc_tcp_ports = 88
     spake_preauth_kdc_challenge = edwards25519
    [realms]
       PUNE.IBM.COM = {
          master_key_type = aes256-cts-hmac-sha384-192
          acl_file = /var/kerberos/krb5kdc/kadm5.acl
          dict_file = /usr/share/dict/words
          default_principal_flags = +preauth
          admin_keytab = /var/kerberos/krb5kdc/kadm5.keytab
          supported_enctypes = aes256-cts-hmac-sha384-192:normal aes128-cts-hmac-sha256-128:normal aes256-cts-hmac-sha1-96:normal aes128-cts-hmac-sha1-96:normal camellia256-cts-cmac:normal camellia128-cts-cmac:normal arcfour-hmac-md5:normal
         # Supported encryption types for FIPS mode:
         #supported_enctypes = aes256-cts-hmac-sha384-192:normal aes128-cts-hmac-sha256-128:normal
      }
  4. Create the KDC database using kdb5_util:

    Note

    Ensure that the host name can be resolved via either DNS or /etc/hosts.

    [root@host ~]# kdb5_util create -s -r [PUNE.IBM.COM](http://pune.ibm.com/)
    
    Initializing database '/var/kerberos/krb5kdc/principal' for realm 'PUNE.IBM.COM',
    master key name 'K/M@PUNE.IBM.COM'
    You will be prompted for the database Master Password.
    It is important that you NOT FORGET this password.
    Enter KDC database master key:
    Re-enter KDC database master key to verify:
  5. Add administrators to the ACL file:

    [root@host ~]# cat /var/kerberos/krb5kdc/kadm5.acl
    
    */admin@PUNE.IBM.COM	*

    The output indicates that any principal in the PUNE.IBM.COM realm with an admin instance has all administrative privileges.

  6. Add administrators to the Kerberos database:

    [root@host ~]# kadmin.local
    
    Authenticating as principal root/admin@PUNE.IBM.COM with password.
    kadmin.local:  addprinc admin/admin@PUNE.IBM.COM
    No policy specified for admin/admin@PUNE.IBM.COM; defaulting to no policy
    Enter password for principal "admin/admin@PUNE.IBM.COM":
    Re-enter password for principal "admin/admin@PUNE.IBM.COM":
    Principal "admin/admin@PUNE.IBM.COM" created.
    kadmin.local:
  7. Start kdc and kadmind:

    # krb5kdc
    # kadmind

Verification

  • Check if kdc and kadmind are running properly:

    # ps -eaf | grep krb
    
    root     27836     1  0 07:35 ?        00:00:00 krb5kdc
    root     27846 13956  0 07:35 pts/8    00:00:00 grep --color=auto krb
    # ps -eaf | grep kad
    root     27841     1  0 07:35 ?        00:00:00 kadmind
    root     27851 13956  0 07:36 pts/8    00:00:00 grep --color=auto kad
  • Check if the setup is correct:

    [root@host ~]# kinit admin/admin
    Password for admin/admin@PUNE.IBM.COM:
    
    [root@ceph-mani-o7fdxp-node4 ~]# klist
    Ticket cache: KCM:0
    Default principal: admin/admin@PUNE.IBM.COM
    
    Valid starting     Expires            Service principal
    10/25/23 06:37:08  10/26/23 06:37:01  krbtgt/PUNE.IBM.COM@PUNE.IBM.COM
        renew until 10/25/23 06:37:08

11.14.2. Setting up the Kerberos client

The Kerberos client machine should be time synced with KDC. Ensure to sync the KDC and clients by using NTP. Time difference of five minutes or more leads to Kerberos authentication failure and throws a clock skew error. This step is a prerequisite on all the systems which are going to participate in Kerberos authentication like NFS clients, hosts where NFS Ganesha containers are going to run.

Procedure

  1. Check the required RPMs

    [root@host ~]# rpm -qa | grep krb5
    
    krb5-libs-1.20.1-9.el9_2.x86_64
    krb5-pkinit-1.20.1-9.el9_2.x86_64
    krb5-workstation-1.20.1-9.el9_2.x86_64
  2. Update the krb5.conf file similar to the one on KDC server:

    [root@host ~]# cat /etc/krb5.conf
    
    # To opt out of the system crypto-policies configuration of krb5, remove the
    # symlink at /etc/krb5.conf.d/crypto-policies which will not be recreated.
    
    includedir /etc/krb5.conf.d/
    [logging]
        default = [FILE:/var/log/krb5libs.log](file:///var/log/krb5libs.log)
        kdc = [FILE:/var/log/krb5kdc.log](file:///var/log/krb5kdc.log)
        admin_server = [FILE:/var/log/kadmind.log](file:///var/log/kadmind.log)
    [libdefaults]
        dns_lookup_realm = false
        ticket_lifetime = 24h
        renew_lifetime = 7d
        forwardable = true
        rdns = false
        pkinit_anchors = [FILE:/etc/pki/tls/certs/ca-bundle.crt](file:///etc/pki/tls/certs/ca-bundle.crt)
        spake_preauth_groups = edwards25519
        dns_canonicalize_hostname = fallback
        qualify_shortname = ""
        default_realm = PUNE.IBM.COM
       default_ccache_name = KEYRING:persistent:%{uid}
    [realms]
      PUNE.IBM.COM = {
         kdc = 10.0.210.222:88
         admin_server = 10.0.210.222:749
      }
    [domain_realm]
      .IBM.com = PUNE.IBM.COM
      IBM.com = PUNE.IBM.COM

Verification

  • Validate the client settings:

    [root@host ~]# kinit admin/admin
    
    Password for admin/admin@PUNE.IBM.COM:
    [root@ceph-mani-o7fdxp-node5 ~]# klist
    Ticket cache: KCM:0
    Default principal: admin/admin@PUNE.IBM.COM
    
    Valid starting     Expires            Service principal
    10/25/23 08:49:12  10/26/23 08:49:08  krbtgt/PUNE.IBM.COM@PUNE.IBM.COM
      renew until 10/25/23 08:49:12

11.14.3. NFS specific Kerberos setup

You need to create service principals for both NFS server and client. The respective keys are stored in the keytab files. These principals are required to set up the initial security context required by GSS_RPCSEC. These service principals have format like nfs/@REALM. You can copy the /etc/krb5.conf file from the working system to the Ceph nodes.

Procedure

  1. Create service principal for that host:

    [root@host ~]# kinit admin/admin
    
    Password for admin/admin@PUNE.IBM.COM:
    [root@host ~]# kadmin
    Authenticating as principal admin/admin@PUNE.IBM.COM with password.
    Password for admin/admin@PUNE.IBM.COM:
    kadmin:  addprinc -randkey nfs/<hostname>.ibm.com
    No policy specified for nfs/<hostname>.ibm.com@PUNE.IBM.COM; defaulting to no policy
    Principal "nfs/<hostname>.ibm.com@PUNE.IBM.COM" created.
  2. Add the key to the keytab file:

    Note

    During this step, you are already on the NFS server and using kadmin interface. Here, the keytab operations reflect on the keytab of the NFS server.

    kadmin:  ktadd nfs/<hostname>.ibm.com
    
    Entry for principal nfs/<hostname>.ibm.com with kvno 2, encryption type aes256-cts-hmac-sha384-192 added to keytab [FILE:/etc/krb5.keytab](file:///etc/krb5.keytab).
    Entry for principal nfs/<hostname>.ibm.com with kvno 2, encryption type aes128-cts-hmac-sha256-128 added to keytab [FILE:/etc/krb5.keytab](file:///etc/krb5.keytab).
    Entry for principal nfs/<hostname>.ibm.com with kvno 2, encryption type aes256-cts-hmac-sha1-96 added to keytab [FILE:/etc/krb5.keytab](file:///etc/krb5.keytab).
    Entry for principal nfs/<hostname>.ibm.com with kvno 2, encryption type aes128-cts-hmac-sha1-96 added to keytab [FILE:/etc/krb5.keytab](file:///etc/krb5.keytab).
    Entry for principal nfs/<hostname>.ibm.com with kvno 2, encryption type camellia256-cts-cmac added to keytab [FILE:/etc/krb5.keytab](file:///etc/krb5.keytab).
    Entry for principal nfs/<hostname>.ibm.com with kvno 2, encryption type camellia128-cts-cmac added to keytab [FILE:/etc/krb5.keytab](file:///etc/krb5.keytab).
    Entry for principal nfs/<hostname>.ibm.com with kvno 2, encryption type arcfour-hmac added to keytab [FILE:/etc/krb5.keytab](file:///etc/krb5.keytab).
    kadmin:
  3. Run steps 1 and 2 on all the Ceph nodes where NFS Ganesha containers are running and all the NFS clients.

11.14.4. NFS Ganesha container settings

Follow the below steps to configure the NFS Ganesha settings in the Ceph environment.

Procedure

  1. Retrieve the existing NFS Ganesha container configuration:

    [ceph: root@host /]# ceph orch ls --service-type nfs --export
    
    service_type: nfs
    service_id: c_ganesha
    service_name: nfs.c_ganesha
    placement:
      hosts:
      - host1
      - host2
      - host3
    spec:
      port: 2049
  2. Modify the container configuration to pass on /etc/krb5.conf and /etc/krb5.keytab` files to the container from the host. These files will be referred by NFS Ganesha at runtime to validate incoming service tickets and secure the communication between Ganesha and NFS client (krb5p).

    [root@host ~]# cat nfs.yaml
    
    service_type: nfs
    service_id: c_ganesha
    service_name: nfs.c_ganesha
    placement:
      hosts:
      - host1
      - host2
      - host3
    spec:
      port: 2049
    extra_container_args:
      - "-v"
      - "/etc/krb5.keytab:/etc/krb5.keytab:ro"
      - "-v"
      - "/etc/krb5.conf:/etc/krb5.conf:ro"
  3. Make the modified nfs.yaml file available inside the cephadm shell:

    [root@host ~]# cephadm shell --mount nfs.yaml:/var/lib/ceph/nfs.yaml
    
    Inferring fsid ff1c1498-73ec-11ee-af38-fa163e9a17fd
    Inferring config /var/lib/ceph/ff1c1498-73ec-11ee-af38-fa163e9a17fd/mon.ceph-msaini-qp49z7-node1-installer/config
    Using ceph image with id 'fada497f9c5f' and tag 'ceph-7.0-rhel-9-containers-candidate-73711-20231018030025' created on 2023-10-18 03:03:39 +0000 UTC
    registry-proxy.engineering.ibm.com/rh-osbs/rhceph@sha256:e66e5dd79d021f3204a183f5dbe4537d0c0e4b466df3b2cc4d50cc79c0f34d75
  4. Validate whether the file has the required changes:

    [ceph: root@host /]# cat /var/lib/ceph/nfs.yaml
    
    service_type: nfs
    service_id: c_ganesha
    service_name: nfs.c_ganesha
    placement:
      hosts:
      - host1
      - host2
      - host3
    spec:
      port: 2049
    extra_container_args:
      - "-v"
      - "/etc/krb5.keytab:/etc/krb5.keytab:ro"
      - "-v"
      - "/etc/krb5.conf:/etc/krb5.conf:ro"
  5. Apply the required changes to NFS Ganesha container and redeploy the container:

    [ceph: root@host /]# ceph orch apply -i /var/lib/ceph/nfs.yaml
    
    Scheduled nfs.c_ganesha update...
    [ceph: root@ceph-msaini-qp49z7-node1-installer /]# ceph orch redeploy nfs.c_ganesha
    Scheduled to redeploy nfs.c_ganesha.1.0.ceph-msaini-qp49z7-node1-installer.sxzuts on host 'ceph-msaini-qp49z7-node1-installer'
    Scheduled to redeploy nfs.c_ganesha.2.0.ceph-msaini-qp49z7-node2.psuvki on host 'ceph-msaini-qp49z7-node2'
    Scheduled to redeploy nfs.c_ganesha.0.0.ceph-msaini-qp49z7-node3.qizzvk on host 'ceph-msaini-qp49z7-node3'
  6. Validate whether the redeployed service has the required changes:

    [ceph: root@host /]# ceph orch ls --service-type nfs --export
    
    service_type: nfs
    service_id: c_ganesha
    service_name: nfs.c_ganesha
    placement:
      hosts:
      - ceph-msaini-qp49z7-node1-installer
      - ceph-msaini-qp49z7-node2
      - ceph-msaini-qp49z7-node3
    extra_container_args:
    - -v
    - /etc/krb5.keytab:/etc/krb5.keytab:ro
    - -v
    - /etc/krb5.conf:/etc/krb5.conf:ro
    spec:
      port: 2049
  7. Modify the export definition to have krb5* (krb5, krb5i, krb5p) security flavor:

    Note

    You can create such an export after completing the above setup.

    [ceph: root@host /]# ceph nfs export info c_ganesha /exp1
    
    {
      "access_type": "RW",
      "clients": [],
      "cluster_id": "c_ganesha",
      "export_id": 1,
      "fsal": {
        "fs_name": "fs1",
        "name": "CEPH",
        "user_id": "nfs.c_ganesha.1"
      },
      "path": "/volumes/_nogroup/exp1/81f9a67e-ddf1-4b5a-9fe0-d87effc7ca16",
      "protocols": [
        4
      ],
      "pseudo": "/exp1",
      "sectype": [
        "krb5"
      ],
      "security_label": true,
      "squash": "none",
      "transports": [
        "TCP"
      ]
    }

11.14.5. NFS Client side actions

Below are some of the operations that NFS clients can undertake.

Procedure

  1. Create service principal:

    kadmin:  addprinc -randkey nfs/<hostname>.ibm.com@PUNE.IBM.COM
    No policy specified for nfs/<hostname>.ibm.com@PUNE.IBM.COM; defaulting to no policy
    Principal "nfs/<hostname>.ibm.com@PUNE.IBM.COM" created.
    kadmin:  ktadd nfs/<hostname>.ibm.com@PUNE.IBM.COM
    Entry for principal nfs/<hostname>.ibm.com@PUNE.IBM.COM with kvno 2, encryption type aes256-cts-hmac-sha384-192 added to keytab [FILE:/etc/krb5.keytab](file:///etc/krb5.keytab).
    Entry for principal nfs/<hostname>.ibm.com@PUNE.IBM.COM with kvno 2, encryption type aes128-cts-hmac-sha256-128 added to keytab [FILE:/etc/krb5.keytab](file:///etc/krb5.keytab).
    Entry for principal nfs/<hostname>.ibm.com@PUNE.IBM.COM with kvno 2, encryption type aes256-cts-hmac-sha1-96 added to keytab [FILE:/etc/krb5.keytab](file:///etc/krb5.keytab).
    Entry for principal nfs/<hostname>.ibm.com@PUNE.IBM.COM with kvno 2, encryption type aes128-cts-hmac-sha1-96 added to keytab [FILE:/etc/krb5.keytab](file:///etc/krb5.keytab).
    Entry for principal nfs/<hostname>.ibm.com@PUNE.IBM.COM with kvno 2, encryption type camellia256-cts-cmac added to keytab [FILE:/etc/krb5.keytab](file:///etc/krb5.keytab).
    Entry for principal nfs/<hostname>.ibm.com@PUNE.IBM.COM with kvno 2, encryption type camellia128-cts-cmac added to keytab [FILE:/etc/krb5.keytab](file:///etc/krb5.keytab).
    Entry for principal nfs/<hostname>.ibm.com@PUNE.IBM.COM with kvno 2, encryption type arcfour-hmac added to keytab [FILE:/etc/krb5.keytab](file:///etc/krb5.keytab).
  2. Restart rpc.gssd service to take effect of the modified/new keytab file:

    # systemctl restart rpc-gssd
  3. Mount the NFS export:

    Syntax

    [root@host ~]# mount -t nfs -o vers=4.1,port=2049 <IP>:/<export_name> >mount_point>

    Example

    mount -t nfs -o vers=4.1,port=2049 10.8.128.233:/ganesha /mnt/test/

  4. Create users. Once the NFS export is mounted, regular users are used to work with mounted exports. These regular users (generally local users on the system or users from centralized system like LDAP/AD) need to be part of Kerberos setup. Based on the kind of setup, local users need to be created in KDC as well.

11.14.6. Validating the setup

Follow the below steps to validate the setup.

Procedure

  • Access the export as normal user, without Kerberos tickets:

    [user@host ~]$ klist
    
    klist: Credentials cache 'KCM:1001' not found
    
    [user@host ~]$ cd /mnt
    -bash: cd: /mnt: Permission denied
  • Access the export as normal user, with Kerberos tickets:

    [user@host ~]$ kinit sachin
    
    Password for user@PUNE.IBM.COM:
    
    [user@host ~]$ klist
    Ticket cache: KCM:1001
    Default principal: user@PUNE.IBM.COM
    
    Valid starting     Expires            Service principal
    10/27/23 12:57:21  10/28/23 12:57:17  krbtgt/PUNE.IBM.COM@PUNE.IBM.COM
      renew until 10/27/23 12:57:21
    
    [user@host ~]$ cd /mnt
    
    [user@host mnt]$ klist
    
    Ticket cache: KCM:1001
    Default principal: user@PUNE.IBM.COM
    
    Valid starting     Expires            Service principal
    10/27/23 12:57:21  10/28/23 12:57:17  krbtgt/PUNE.IBM.COM@PUNE.IBM.COM
    renew until 10/27/23 12:57:21
    10/27/23 12:57:28  10/28/23 12:57:17  nfs/ceph-msaini-qp49z7-node1-installer.ibm.com@
    renew until 10/27/23 12:57:21
    Ticket server: nfs/ceph-msaini-qp49z7-node1-installer.ibm.com@PUNE.IBM.COM
Note

Note: Tickets for nfs/ service are observed on the client.

Red Hat logoGithubRedditYoutubeTwitter

Apprendre

Essayez, achetez et vendez

Communautés

À propos de la documentation Red Hat

Nous aidons les utilisateurs de Red Hat à innover et à atteindre leurs objectifs grâce à nos produits et services avec un contenu auquel ils peuvent faire confiance.

Rendre l’open source plus inclusif

Red Hat s'engage à remplacer le langage problématique dans notre code, notre documentation et nos propriétés Web. Pour plus de détails, consultez leBlog Red Hat.

À propos de Red Hat

Nous proposons des solutions renforcées qui facilitent le travail des entreprises sur plusieurs plates-formes et environnements, du centre de données central à la périphérie du réseau.

© 2024 Red Hat, Inc.