Feat(Backup): Add native support for backup to Azure.

[rohanprasad]

Fixes DGRAPH-3384

Currently, we use MinIO to write backup files to Azure blob storage. To do this we need to start a minio server (with Azure credentials) and trigger a backup with the destination as the minio server. This requires the user to run the minio server every time (or always) when a backup needs to be triggered. Using native SDK will help get rid of the extra service.

Azure storage (AZS) requires the user to get secret. When starting an alpha we can set env variable:

AZURE_STORAGE_KEY="secret_key_here" dgraph alpha

to allow access to AZS. To trigger a backup the user can use the admin endpoint to initiate the backup. The destination should be set as:

"https://<account_name>.blob.core.windows.net//path/to/dest"

---

This change is 

[ahsanbarkati]

Reviewable status: 0 of 3 files reviewed, 4 unresolved discussions (waiting on @manishrjain, @rohanprasad, and @vvbalaji-dgraph)

---

x/handlers.go, line 100 at r1 (raw file):

	}

	if strings.HasSuffix(uri.Host, "blob.core.windows.net") {

Is this the right way? Why are we not using uri.Scheme?

---

x/handlers.go, line 371 at r1 (raw file):

// Helper function to get the account, container and path of the destination folder from an Azure
// URL.
func getAzDetailsFromUri(uri *url.URL) (accountName, bucketName, pathName string, err error) {

Don't use named return values. They are confusing. Also avoid having more than 2 return values.
You can pack them in a struct.

---

x/handlers.go, line 375 at r1 (raw file):

	parts := strings.Split(uri.Host, ".")
	if len(parts) > 0 {
		accountName = parts[0]

Wouldn't parts[0] be https://<account_name>. Do we want the https part as well?

---

x/handlers.go, line 381 at r1 (raw file):

	}

	parts = strings.Split(uri.Path, string(filepath.Separator))

Assuming path is https://<account_name>.... Wouldn't parts[1] contain an empty string?

[rohanprasad]

---

x/handlers.go, line 100 at r1 (raw file):

Previously, ahsanbarkati (Ahsan Barkati) wrote…

Is this the right way? Why are we not using uri.Scheme?

Haven't found a scheme-based URL for Azure blob storage. Assuming an https path here:
https://<account_name>.blob.core.windows.net//path/to/object (added in comments as well)

[rohanprasad]

---

x/handlers.go, line 371 at r1 (raw file):

Previously, ahsanbarkati (Ahsan Barkati) wrote…

Don't use named return values. They are confusing. Also avoid having more than 2 return values.
You can pack them in a struct.

Will move them in a struct.

[rohanprasad]

---

x/handlers.go, line 375 at r1 (raw file):

Previously, ahsanbarkati (Ahsan Barkati) wrote…

Wouldn't parts[0] be https://<account_name>. Do we want the https part as well?

Host doesn't contains scheme, so this should work well.

[rohanprasad]

---

x/handlers.go, line 381 at r1 (raw file):

Previously, ahsanbarkati (Ahsan Barkati) wrote…

Assuming path is https://<account_name>.... Wouldn't parts[1] contain an empty string?

Path is of the form: /container/path/to/object. When split, parts[0] will be empty and parts[1] will contain the container/bucket name.

[ahsanbarkati]

LGTM. Please test these APIs manually.