Stateful application storage operations
With Portworx, you can backup stateful applications and their volumes to an external object store and restore to the same cluster or to a new cluster. When it comes to timing, you can perform backup operations manually whenever you wish, or you can schedule them with a schedulePolicy. You can even use Portworx's backup functionality to clone applications between namespaces by restoring to a different namespace in the same cluster.
Prerequisites
All application storage operations require the following prerequisites:
- Administrator-level cluster privileges if you operate between namespaces
Configure Application Backups
Configure application backups by creating and applying the following CRDs:
- backupLocation
- applicationBackup
Create a backupLocation CRD
Use the backupLocation CRD to specify the configuration information for your object store. Portworx supports the following object stores:
- Any S3-compliant object store
- Azure Blob Storage
- Google Cloud Storage
The example in this document uses an S3 bucket to store application backup data. To create backupLocations using other object stores, see the Stateful application CRD reference section of the documentation.
"endpoint" : "bucketEndpoint.com"
"key": "ABCDEF1234567890"
"secret": "ABCDEF1234567890ABCDEF1234567890ABCDEF1234567890"
If you're restoring backups to a different cluster from the one you took your backups from, you must create a backupLocation CRD on your new cluster that matches the backupLocation CRD on your original cluster.
If you want to use S3 Object Lock, refer to S3 Object Lock in PX-Backup.
You can specify your object store credentials directly in the BackupLocation configuration as plaintext or use a Kubernetes secret:
Plaintext credentials
- Create a backupLocation YAML file, specifying the following:
- name: the backupLocation object's name
- namespace: the namespace the backupLocation exists in
- location:
-
type: the object store type
-
path: the bucket Portworx will use for the backup
-
sync: If you're restoring to a new cluster, set
sync
to true to allow your new cluster to retrieve the previous backups from your backup location. -
s3Config:
-
region: which region your s3 bucket is located in
-
accessKeyID: your bucket's accessKeyID
-
secretAccessKey: your bucket's secretAccessKey
-
endpoint: the URL or IP address of your bucket
-
disableSSL: whether or not to disable SSL
apiVersion: stork.libopenstorage.org/v1alpha1
kind: BackupLocation
metadata:
name: mysql
namespace: mysql-app
annotations:
stork.libopenstorage.org/skipresource: "true"
location:
type: s3
path: "bucket-name"
sync: true
s3Config:
region: us-east-1
accessKeyID: XXXX
secretAccessKey: XXXX
endpoint: "https://bucketEndpoint.com"
disableSSL: false
noteIf you use URL as your bucket endpoint, you must include the http prefix: either
https://
orhttp://
, depending on whether or not you're using SSL. -
-
-
Apply the YAML:
kubectl apply -f backLo.yaml
Kubernetes secret containing your credentials
- Create a Kubernetes Secret YAML file, specifying the following:
- name: the Secret object's name
- namespace: the namespace the Secret exists in
- stringData:
-
region: which region your s3 bucket is located in
-
accessKeyID: your bucket's accessKeyID
-
secretAccessKey: your bucket's secretAccessKey
-
endpoint: the URL or IP address of your bucket
-
disableSSL: whether or not to disable SSL
-
encryptionKey: your secret's encryption key
apiVersion: v1
kind: Secret
metadata:
name: s3secret
namespace: mysql
annotations:
stork.libopenstorage.org/skipresource: "true"
stringData:
region: us-east-1
accessKeyID: <access-key>
secretAccessKey: <secret-key>
endpoint: "X.X.X.141:9010"
disableSSL: "false"
encryptionKey: "testKey"
-
-
Apply the Secret's YAML file:
kubectl apply -f s3secret.yaml
-
Create a backupLocation YAML file, specifying the following:
- name: the backupLocation object's name
- namespace: the namespace the backupLocation exists in
- location:
-
type: the object store type
-
path: the bucket Portworx will use for the backup
-
secretConfig: the Secret object containing your bucket's credentials
-
sync: If you're restoring to a new cluster, set
sync
to true to allow your new cluster to retrieve the previous backups from your backup location.apiVersion: stork.libopenstorage.org/v1alpha1
kind: BackupLocation
metadata:
name: mysql-backup
namespace: mysql
annotations:
stork.libopenstorage.org/skipresource: "true"
location:
type: s3
path: "bucket-name"
secretConfig: s3secret
sync: true
-
If you use URL as your bucket endpoint, you must include the http prefix: either https://
or http://
, depending on whether or not you're using SSL.
-
Apply the YAML:
kubectl apply -f backLo.yaml
Create an applicationBackup CRD
Use the applicationBackup CRD to specify what namespaces have their applications backed-up.
- Create an applicationbackup YAML file, specifying the following:
-
name: the applicationBackup object's name
-
namespace: the namespace the applicationBackup exists in
-
spec:
- backupLocation: what backupLocation object to use to determine where to send the backup
- namespaces: the namespaces to backup
- reclaimPolicy: what happens to objects in the object store when the
ApplicationBackup
object is deleted, eitherDelete
orRetain
- selectors: define specific labels to determine which objects and volumes are backed-up
- preExecRule: what rule to run before performing backup
- postExecRule: what rule to run after performing backup
apiVersion: stork.libopenstorage.org/v1alpha1
kind: ApplicationBackup
metadata:
name: backup
namespace: mysql-app
spec:
backupLocation: mysql
namespaces:
- mysql-app
reclaimPolicy: Delete
selectors:
preExecRule:
postExecRule:
-
Apply the YAML:
kubectl apply -f appBack.yaml
-
Verify the applicationBackup object's status with the
storkctl get
command. You can see what stage the backup is in, as well as its status:storkctl get applicationbackup -n mysql-app
NAME STAGE STATUS VOLUMES RESOURCES CREATED ELAPSED
backup Volumes Pending 0/0 0 11 Sep 19 22:10 UTC 19.023065649sYou can also describe the object to get more information about the backup. This is useful for troubleshooting:
kubectl describe applicationbackup.stork.libopenstorage.org -n mysql-app
The following error message displays when your credentials are invalid.
Warning Failed 7s stork Error starting ApplicationBackup for volumes: rpc error: code = Internal desc = Failed to create backup: Failed to validate credentials: error validating credential: AccessDenied: Access Denied.
status code: 403, request id: 15C38376E0A0EA7F, host id:
Create an ApplicationBackupSchedule CRD
The ApplicationBackupSchedule CRD associates a SchedulePolicy with an application backup operation, allowing you to schedule when and how application backups are performed.
-
Create a file named
backupSchedulePolicy.yaml
with the following content:apiVersion: stork.libopenstorage.org/v1alpha1
kind: SchedulePolicy
metadata:
name: backupSchedule
policy:
interval:
intervalMinutes: 60
retain: 5
daily:
time: "10:14PM"
retain: 5
weekly:
day: "Thursday"
time: "10:13PM"
retain: 5
monthly:
date: 14
time: "8:05PM"
retain: 5For details about how you can configure a schedule policy, see the SchedulePolicy page.
-
Apply the spec by entering the following command:
kubectl apply -f backupSchedulePolicy.yaml
-
Display your schedule policy. Enter the
storkctl get
command passing it the name of your policy:storkctl get schedulepolicy
NAME INTERVAL-MINUTES DAILY WEEKLY MONTHLY
testpolicy 60 10:14PM Thursday@10:13PM 14@8:05PM -
Create an ApplicationBackupSchedule YAML file, specifying the following:
-
name: the applicationBackupSchedule object's name
-
namespace: the namespace the applicationBackupSchedule exists in
-
spec.schedulePolicyName: the name of the schedule policy that defines when backup actions happen
-
spec.template.spec.backupLocation: the name of the backup location spec
-
spec.template.spec.namespaces: namespaces which will be backed up
-
spec.template.spec.reclaimPolicy: what happens to objects in the object store when the
ApplicationBackup
object is deletedapiVersion: stork.libopenstorage.org/v1alpha1
kind: ApplicationBackupSchedule
metadata:
name: backup
namespace: mysql
spec:
schedulePolicyName: testpolicy
template:
spec:
backupLocation: mysql
namespaces:
- mysql
reclaimPolicy: Delete
Restore an application
You can restore an application by applying an ApplicationRestore object.
If you're restoring an application across namespaces on OpenShift, you must modify your destination namespace to include the same supplemental group annotation values as your source namespace:
annotations:
openshift.io/sa.scc.mcs: s0:c26,c25
openshift.io/sa.scc.supplemental-groups: 1001990000/10000
openshift.io/sa.scc.uid-range: 1001990000/10000
- Create an ApplicationRestore YAML file, specifying the following:
- name: the ApplicationRestore object's name
- namespace: the ApplicationRestore object's namespace
- spec.backupName: the name of the
applicationBackup
object to restore from - spec.backupLocation: which backup location object to get application backups from
- spec.namespaceMapping: a map of source and destination namespaces, allowing you to restore a backup to a different namespace
- spec.replacePolicy: specifies whether you want to delete or retain any matching existing resource in the target namespace
apiVersion: stork.libopenstorage.org/v1alpha1
kind: ApplicationRestore
metadata:
name: restore
namespace: mysql-app
spec:
backupName: backup
backupLocation: mysql
namespaceMapping:
<backup_namespace>: <restore_namespace>
replacePolicy: Delete
You can run the storkctl get applicationbackup -n namespace
command to help you see which backup to restore from.
-
Apply the YAML:
kubectl apply -f appRestore.yaml
The moment you apply the YAML, the application begins to restore. Monitor the status with the storkctl get
command:
storkctl get applicationrestore -n mysql-app
NAME STAGE STATUS VOLUMES RESOURCES CREATED ELAPSED
restore Final Successful 1/1 3 11 Sep 19 23:32 UTC 35s
Verify its status with the kubectl get pods
command:
kubectl get pods -n mysql-app
NAME READY STATUS RESTARTS AGE
mysql-6d69b99774-2bv4m 0/1 Pending 0 3s
Clone an Application
You can clone an application to a different namespace. You must create the ApplicationClone
object in the admin namespace, which is kube-system
by default.
-
If you're cloning an application across namespaces on OpenShift, you must modify your destination namespace to include the same supplemental group annotation values as your source namespace:
annotations:
openshift.io/sa.scc.mcs: s0:c26,c25
openshift.io/sa.scc.supplemental-groups: 1001990000/10000
openshift.io/sa.scc.uid-range: 1001990000/10000 -
Distributed apps, such as Cassandra, may use the same node IDs on the destination namespace as their source, causing disruption when the new nodes join the source cluster.
-
Create a an ApplicationClone YAML file, specifying the following:
-
name: the ApplicationClone object's name
-
namespace: the ApplicationClone object's namespace
-
spec.sourceNamespace: the namespace you want to clone applications from
-
spec.destinationNamespace: the namespace you want to clone applications to
apiVersion: stork.libopenstorage.org/v1alpha1
kind: ApplicationClone
metadata:
name: clone-mysql
namespace: kube-system
spec:
sourceNamespace: mysql-app
destinationNamespace: clone-mysql
-
-
Apply the YAML file:
kubectl apply -f appClone.yaml