Configuring Persistent Disk Parameters¶
Before reading this section, complete the steps in Before You Begin.
To create a custom deployment, you need to configure
persistent storage that Pachyderm will use to store metadata.
You can do so by using the
that creates a PersistentVolume (PV) backend on a supported provider.
Pachyderm has automated configuration for styles of backend for the following major cloud providers:
- Amazon Web Services™ (AWS)
- Google Cloud Platform™ (GCP)
- Microsoft® Azure™
Choosing one of these providers creates a configuration close to what you need. After carefully reading the section below, consult with your Kubernetes administrators on which provider to choose. You might need to then edit your manifest manually, based on configuration information they provide to you.
For each of the providers above, the final configuration depends on which of the following flags you define:
--dynamic-etcd-nodesflag is used when your Kubernetes installation is configured to use StatefulSets. Many Kubernetes deployments use StatefulSets as a reliable solution that ensures the persistence of pod storage. Your on-premises Kubernetes installation might also be configured to use StatefulSets. The
--dynamic-etcd-nodesflag specifies the number of
etcdnodes that your deployment creates. Pachyderm recommends that you keep this number at
1. If you want to change it, consult with your Pachyderm support team. This flag creates a
StatefulSetthat uses the standard
NOTE Consult with your Kubernetes administrator about the StorageClass that you should use for
etcdin your Kubernetes deployment. If you need to use a different than the default setting, you can use the
--etcd-storage-classflag to specify the StorageClass.
--static-etc-volumeflag is used when your Kubernetes installation has not been configured to use StatefulSets. When you specify
--static-etc-volumeflag, Pachyderm creates a static volume for
etcd. Pachyderm creates a PV with a spec appropriate for each of the cloud providers:
aws: awsElasticBlockStore for Amazon Web Services
azure: azureDisk for Microsoft Azure
As stated above, the specifics of one of these choices might not match precisely what your on-premises deployment requires. To determine the closest correct choices for your on-prem infrastructure, consult with your Kubernetes administrators. You might have to then edit your manifest manually, based on configuration information they provide to you.
Example invocation with persistent disk parameters¶
This example on-premises cluster has StatefulSets enabled, with the standard etcd storage class configured. The deployment command uses the following flags:
pachctl deploy custom --persistent-disk aws --object-store <object store backend> \ any-string 10 \ <object store arg 1> <object store arg 2> <object store arg 3> <object store arg 4> \ --dynamic-etcd-nodes 1 [optional flags]
--persistent-disk flag takes two arguments
that you specify right after the single argument to the
Although the first argument is required, Pachyderm ignores it.
Therefore, you can set it to any text value, such as
any-string in the
The second argument is the size,
in gigabytes (GB), that Pachyderm requests for the
A good value for most deployments is 10.
After completing the steps described in this section, proceed to Configuring Object Store.