Here are instructions to help you read and write to a mounted file system from your apps. In VMware Tanzu Application Service for VMs (TAS for VMs), a volume service provides a volume so your app can read or write to a reliable, persistent file system.
NoteIf you are using VMware Tanzu Application Service for VMs [Windows], see Using SMB Volumes in .NET Apps.
NFS is not available on Windows systems.
Before you can use a volume service with your app, find out if any volume services are available for your app.
Log in to the Cloud Foundry Command Line Interface (cf CLI). Run:
cf login
List available NFS volume services. Run:
cf marketplace
See the following example output of the NFS volume service:
cf marketplace service plans description nfs Existing Service for connecting to NFS volumes
Do one of the following:
The following sections describe how to mount an external file system to your app.
To use a volume service deployed by your Tanzu Operations Manager admin, you must first create an instance of the specific volume service that you need.
Note You can also bind volume services using an app manifest. However, app manifests do not support bind configuration. To bind a volume service using an app manifest, you must specify bind configuration when you create the service instance. The releases that support this are nfs-volume
v1.3.1 and later and smb-volume
v1.0.0 and later. For more information, see Services in Deploying with App Manifests.
To create and bind an instance for the volume service:
Create a service instance:
Bind your service instance to an app:
cf bind-service YOUR-APP SERVICE-NAME -c '{"mount":"OPTIONAL-MOUNT-PATH","readonly":true}'
The app file operations use the UID of the running app process. For buildpack apps, this UID is always 2000
. For Docker apps, the effective UID is the same as the UID of the process inside the Docker container, except for root
, which is mapped to 4294967294
outside the Docker container.
Or to use UID/GID mapping:
cf bind-service YOUR-APP SERVICE-NAME -c '{"uid":"UID","gid":"GID","mount":"OPTIONAL-MOUNT-PATH","readonly":true}'
Where:
YOUR-APP
is the name of the app for which you want to use the volume service.SERVICE-NAME
is the name of the volume service instance you created in an earlier step.(Optional) UID
and GID
are the UID and GID to use when mounting the share to the app. The GID
and UID
must be positive integer values greater than 0
. Provide the UID and GID as a JSON string in-line or in a file. If you omit uid
and gid
, the driver skips mapfs
mounting and performs only the normal kernel mount of the NFS file system without the performance overhead associated with FUSE mounts.
The key advantage of specifying UID
and GID
is that you can specify different values for different apps, so file permissions can be granted at the app level. If this is not needed, the you can eliminate the performance overhead of mapfs
by managing permissions on the NFS server.
The user specified by uid
must have access to the files on the share. When uid
and gid
are omitted, the app file operations use the UID of the running app process. For buildpack apps, this UID is always 2000
. For Docker apps, the effective UID is the same as the UID of the process inside the Docker container, except for root
, which is mapped to 4294967294
outside the Docker container.
Caution Specifying UID and GID values affects performance because the FUSE file system mapfs is used to translate UID and GID values.
OPTIONAL-MOUNT-PATH
is a JSON string that indicates that the volume must be mounted to a particular path in your app rather than the default path. Choose a path with a root-level directory that already exists in the container, such as /home
, /usr
, or /var
. Important Do not specify a MOUNT-PATH
in the /app
directory, which is where TAS for VMs unpacks the droplet. For more information, see Mount a shared volume in the /app directory.
"readonly":true
is a JSON string that creates a read-only mount. By default, Volume Services mounts a read-write file system. For read-only mounts, the driver activates attribute caching. This results in fewer attribute RPCs and better performance. The following example shows binding my-app
to the nfs_service_instance
and specifying a read-only volume to be mounted to /var/volume1
, passing an in-line JSON string:
cf bind-service my-app nfs_service_instance -c '{"uid":"1000","gid":"1000","mount":"/var/volume1","readonly":true}'
LDAP-only: Bind an LDAP service to an app. Run:
cf bind-service YOUR-APP SERVICE-NAME -c '{"username":"USERNAME","password":"PASSWORD","mount":"OPTIONAL-MOUNT-PATH","readonly":true}'
Where:
YOUR-APP
is the name of the app for which you want to use the volume service.SERVICE-NAME
is the name of the volume service instance you created in an earlier step.USERNAME
and PASSWORD
are the user name and password for the LDAP server. If you omit username
and password
, the driver skips mapfs
mounting and performs only the normal kernel mount of the NFS file system without the overhead associated with FUSE mounts. OPTIONAL-MOUNT-PATH
is a JSON string that indicates the volume must be mounted to a particular path within your app rather than the default path. Choose a path with a root-level directory that already exists in the container, such as /home
, /usr
, or /var
. Important Do not specify a MOUNT-PATH
within the /app
directory, which is where TAS for VMs unpacks the droplet. For more information, see Mount a shared volume in the /app directory.
"readonly":true
is a JSON string that creates a read-only mount. By default, Volume Services mounts a read write file system. For read-only mounts, the driver activates attribute caching. This results in fewer attribute RPCs and better performance. Restage your app. Run:
cf restage YOUR-APP
Where YOUR-APP
is the name of the app.
To access the volume service from your app, you must know which file path to use in your code.
You can view the file path in the details of the service binding, which are available from the VCAP_SERVICES
environment variable. See VCAP_SERVICES.
To access the volume service from your app:
View environment variables for your app. Run:
cf env YOUR-APP
Where YOUR-APP
is the name of your app.
The following is example output of the cf env
command:
$ cf env YOUR-APP "VCAP_SERVICES": { "nfs": [ { "credentials": {}, "label": "nfs", "name": "nfs_service_instance", "plan": "Existing", "provider": null, "syslog_drain_url": null, "tags": [ "nfs" ], "volume_mounts": [ { "container_dir": "/var/vcap/data/153e3c4b-1151-4cf7-b311-948dd77fce64", "device_type": "shared", "mode": "rw" } ] } ] }
Use the properties under volume_mounts
for any information your app needs.
Property | Description |
---|---|
container_dir |
String containing the path to the mounted volume that you bound to your app. |
device_type |
The NFS volume release. This only supports shared devices. A shared device represents a distributed file system that can mount on all app instances simultaneously. |
mode |
String that informs what type of access your app has to NFS, either read-only, ro , or read and write, rw . |
If you specify a mount inside the /app
directory, the app might fail to start and parts of the app droplet can be written to the remote file share. This is because TAS for VMs mounts the volume before moving your compiled app into the droplet.
If your app requires the shared volume to be placed within the /app
directory, you can use a symbolic link at app startup.
To mount a volume in the /app
directory:
Specify a mount volume in a location outside of the /app
directory.
Create a symbolic link at app startup time, prior to launching the app. For example, run:
cf push YOUR-APP -c "ln -s /var/volume1 /app/volume1 && \$HOME/boot.sh"
Where YOUR-APP
is the name of the app.
This section describes how to use the NFS volume service.
TAS for VMs offers two NFS volume services:
nfs
: This volume service provides support for NFS volumes using both v3 and v4.x protocols.
nfs-legacy
(deprecated): Although it is deprecated, this volume service is still available due to the difficulty of retiring services. If you use this service, it performs exactly the same mount as the nfs
service. For information about migrating to nfs
, see Migrate nfs-legacy
Services to nfs
.
Both services offer a single plan called Existing
.
Note NFS is not available on Windows systems.
To create an NFS volume service using the Existing
plan of the nfs
service:
Create an NFS volume service. Run:
cf create-service nfs Existing SERVICE-INSTANCE-NAME -c '{"share":"SERVER/SHARE", "version":"NFS-PROTOCOL"}'
Where:
SERVICE-INSTANCE-NAME
is a name you provide for this NFS volume service instance.SERVER/SHARE
is the NFS address of your server and share. Important Omit the :
that usually follows the server name in the address.
NFS-PROTOCOL
is the NFS protocol you want to use. For example, to use NFSv4, set the version to 4.1
. Valid values are 3
, 4.0
, 4.1
or 4.2
. If you do not specify a version
, the protocol version used is negotiated between client and server at mount time. This usually causes the latest available version to be used. Important Tanzu Application Service versions shipping with nfs-volume versions v7.1.45 - v7.1.47 or v5.0.55 - 5.0.58
as listed in the TAS for VMs Release notes or Isolation Segment Release notes do not support specifying NFS version: 3.0
. These updated the contained nfs-utils (a dependency of nfs-volume-service) to a newer version that uses stricter option parsing.
NFSv3 does not utilize a MINOR version, but NFSv4 introduced MINOR versions that can be specified.
This has been mitigated by adding auto-correction logic to the nfsdriver process available with nfs-volume >= v7.1.48
and >= v5.0.59
.
Confirm that the NFS volume service appears in your list of services. Run:
cf services
With the release of NFS Volume Service v1.5.4, the original fuse-based NFS service is deprecated in favor of the later kernel mount-based NFS service. Existing NFS volume service bindings are listed as nfs-legacy
.
To migrate from nfs-legacy
to the later nfs
service, VMware recommends that you recreate and re-bind your nfs
service instances.
With the release of NFS Volume Service v2.0.0, the nfs-legacy
service uses the nfs
service. To avoid being affected when the nfs-legacy
service is retired, recreate and re-bind your service instances using the nfs
service.
This section describes how to deploy a sample app and bind it to the NFS volume service.
To deploy and bind a sample app:
Clone the GitHub repository for the sample app into your workspace by running these commands:
cd ~/workspace
git clone https://github.com/cloudfoundry/persi-acceptance-tests.git
Change into the persi-acceptance-tests/assets/pora/
directory:
cd ~/workspace/persi-acceptance-tests/assets/pora
Push the pora
test app by running:
cf push pora --no-start
Bind the service to your app. Run:
cf bind-service pora SERVICE-INSTANCE-NAME -c '{"uid":"UID","gid":"GID"}'
Where:
SERVICE-INSTANCE-NAME
: The name of the volume service instance you created previously.UID
and GID
: The UID
and GID
to use when mounting the share to the app. The NFS driver uses these values in the following ways:
UID
and GID
values. UID
and GID
back to the running user UID and default GID. This allows you to interact with your NFS server as a specific user while allowing TAS for VMs to run your app as an arbitrary user. UID
and GID
must be positive integer values. Important In NFS v2.0.0 and later, uid
and gid
values of 0
are no longer permissible because of security concerns.
mount
: Use this option to specify the path at which volumes mount to the application container. The default is an arbitrarily-named directory in /var/vcap/data
. You can edit this value if your app has specific requirements. For example: cf bind-service pora myVolume -c '{"mount":"/var/path"}'
readonly
: When you run the cf bind-service
command, Volume Services mounts a read-write file system by default. You can specify a read-only mount by adding "readonly":true
to the bind configuration JSON string. cache
: When you run the cf bind-service
command, Volume Services mounts the remote file system with attribute caching deactivated by default. You can activate attribute caching using default values by adding "cache":true
to the bind configuration JSON string. Start the app. Run:
cf start pora
Confirm the app is running. Run:
curl http://pora.YOUR-CF-DOMAIN.com
The command returns an instance index for your app.
Confirm the app can access the shared volume. Run:
curl http://pora.YOUR-CF-DOMAIN.com/write
The command writes a file to the share and then reads it back out again.
This section describes using the NFS volume service.
If your Operations Manager deployment has LDAP activated, you can configure LDAP credentials for your NFS Volume Service instance.
To configure LDAP credentials while creating your NFS Volume Service instance:
Specify values for username
and password
in the JSON string for your cf create-service
command:
cf create-service nfs PLAN SERVICE-INSTANCE-NAME -c '{"share":"SERVER/SHARE", "username":"USERNAME", "password":"PASSWORD"}'
Where:
PLAN
is the name of the service plan.SERVICE-INSTANCE-NAME
is a name you provide for this NFS Volume Service instance.SERVER/SHARE
is the NFS address of your server and share.USERNAME
is a user name you provide for this NFS Volume Service instance.PASSWORD
is a password you provide for this NFS Volume Service instance.As of nfs-volume-release
v1.3.1, you can specify bind parameters in advance, when you create a service instance. Use this option if you bind the service to your app in an app manifest, where bind configuration is not supported.
Apps that use file locking through UNIX system calls such as flock()
and fcntl()
or script commands such as flock
to use the nfs
service. The nfs-legacy
service uses a fuse mounting process that does not enforce locks across Diego Cells.
The mapfs UID mapping layer used by the NFS service does not support hard link operations. You get a Function not implemented
error if you try to create a hard link in an NFS share when uid
or username
is specified for the service.
Workarounds for this issue:
ln -s
, instead of hard links.uid
and gid
or the username
and password
parameters to mount the share without UID mapping. For this workaround, the app user must have access to the files on the share.This section describes how to use a Server Message Block (SMB) volume service. For more information about SMB volume services, see the Microsoft documentation in the Microsoft documentation.
NoteIf you are using VMware Tanzu Application Service for VMs [Windows], see Using SMB Volumes in .NET Apps.
TAS for VMs offers an smb
volume service. This volume service provides support for existing SMB shares.
The service offers a single plan called Existing
.
To create an SMB volume service:
Create the service by running:
cf create-service smb Existing SERVICE-INSTANCE-NAME -c '{"share":"//SERVER/SHARE", "version":"SMB-VERSION"}'
Where:
SERVICE-INSTANCE-NAME
is a name you provide for this SMB volume service instance.//SERVER/SHARE
is the SMB address of your server and share.SMB-VERSION
is the SMB protocol version you want to use. For example, to use SMB 2.1, set the version to 2.1
. Valid values are 1.0
, 2.0
, 2.1
, or 3.0
. If you do not specify a version
, the client and server negotiate a protocol version at mount time. The client and server usually select the latest available version. Confirm that the SMB volume service appears in your list of services. Run:
cf services
This section describes how to deploy a sample app and bind it to the SMB volume service.
To deploy and bind a sample app:
Clone the GitHub repository for the sample app into your workspace. Run:
cd ~/workspace
git clone https://github.com/cloudfoundry/persi-acceptance-tests.git
Change into the persi-acceptance-tests/assets/pora/
directory:
cd ~/workspace/persi-acceptance-tests/assets/pora
Push the pora
test app. Run:
cf push pora --no-start
Bind the service to your app. Run:
cf bind-service pora SERVICE-INSTANCE-NAME -c '{"username":"USERNAME","password":"PASSWORD"}'
Where:
SERVICE-INSTANCE-NAME
: The name of the volume service instance you created previously.
USERNAME
and PASSWORD
: The user name and password to use when mounting the share to the app. This allows you to interact with your SMB server as a specific user while allowing TAS for VMs to run your app as an arbitrary user.
(Optional) mount
: Use this option to specify the path at which volumes mount to the application container. The default is an arbitrarily-named directory in /var/vcap/data
. You can edit this value if your app has specific requirements. For example, run:
cf bind-service pora myVolume -c '{"username":"some-user","password":"some-password","mount":"/var/path"}'
(Optional) readonly
: When you run the cf bind-service
command, Volume Services mounts a read-write file system by default. You can specify a read-only mount by adding "readonly":true
to the bind configuration JSON string.
(Optional) domain
: If you use a Windows domain, you can specify a domain
parameter.
Start the app. Run:
cf start pora
Confirm the app is running. Run:
curl http://pora.YOUR-CF-DOMAIN.com
The command returns an instance index for your app.
Confirm the app can access the shared volume. Run:
curl http://pora.YOUR-CF-DOMAIN.com/write
The command writes a file to the share and then reads it back out again.