[pbs-devel] [PATCH proxmox-backup v13 03/26] pbs-api-types: add backing-device to DataStoreConfig

[Hannes Laimer]

On Sun Nov 17, 2024 at 8:27 PM CET, Thomas Lamprecht wrote:
> Am 13.11.24 um 16:00 schrieb Hannes Laimer:
>> Signed-off-by: Hannes Laimer <h.laimer at proxmox.com>
>> ---
>> changes since v12:
>>  * clearify/improve description of `DATASTORE_DIR_NAME_SCHAME`
>> 
>>  pbs-api-types/src/datastore.rs | 31 ++++++++++++++++++++++++++++---
>>  1 file changed, 28 insertions(+), 3 deletions(-)
>> 
>> diff --git a/pbs-api-types/src/datastore.rs b/pbs-api-types/src/datastore.rs
>> index a5704c93..f6c255d3 100644
>> --- a/pbs-api-types/src/datastore.rs
>> +++ b/pbs-api-types/src/datastore.rs
>> @@ -42,7 +42,7 @@ const_regex! {
>>  
>>  pub const CHUNK_DIGEST_FORMAT: ApiStringFormat = ApiStringFormat::Pattern(&SHA256_HEX_REGEX);
>>  
>> -pub const DIR_NAME_SCHEMA: Schema = StringSchema::new("Directory name")
>> +pub const DATASTORE_DIR_NAME_SCHEMA: Schema = StringSchema::new("Either the absolute path to the datastore directory, or a relative on-device path for removable datastores.")
>>      .min_length(1)
>>      .max_length(4096)
>>      .schema();
>> @@ -160,6 +160,9 @@ pub const PRUNE_SCHEMA_KEEP_YEARLY: Schema =
>>          .minimum(1)
>>          .schema();
>>  
>> +/// Base directory where datastores are mounted
>> +pub const DATASTORE_MOUNT_DIR: &str = "/mnt/datastore";
>> +
>>  #[api]
>>  #[derive(Debug, Default, Copy, Clone, PartialEq, Eq, Serialize, Deserialize)]
>>  #[serde(rename_all = "lowercase")]
>> @@ -234,7 +237,7 @@ pub const DATASTORE_TUNING_STRING_SCHEMA: Schema = StringSchema::new("Datastore
>>              schema: DATASTORE_SCHEMA,
>>          },
>>          path: {
>> -            schema: DIR_NAME_SCHEMA,
>> +            schema: DATASTORE_DIR_NAME_SCHEMA,
>>          },
>>          "notify-user": {
>>              optional: true,
>> @@ -273,6 +276,12 @@ pub const DATASTORE_TUNING_STRING_SCHEMA: Schema = StringSchema::new("Datastore
>>              format: &ApiStringFormat::PropertyString(&MaintenanceMode::API_SCHEMA),
>>              type: String,
>>          },
>> +        "backing-device": {
>> +            description: "The UUID of the filesystem partition for removable datastores.",
>> +            optional: true,
>> +            format: &proxmox_schema::api_types::UUID_FORMAT,
>
> FWIW, I get an error about this regex not matching for a USB pen drive I'm testing.
>
> The POST data is:
>
> {
> 	"name": "samsung-stick-foo",
> 	"path": "foo",
> 	"backing-device": "64A5-F009",
> 	"gc-schedule": "daily",
> 	"prune-schedule": "daily",
> 	"comment": "",
> 	"notification-mode": "notification-system"
> }
>
>
> The data of the usb disk I selected
>
> {
> 	"3": {
> 		"devpath": "/dev/sdd",
> 		"disk-type": "hdd",
> 		"gpt": false,
> 		"model": "Flash_Drive_FIT",
> 		"name": "sdd",
> 		"partitions": [
> 			{
> 				"devpath": "/dev/sdd1",
> 				"filesystem": "exfat",
> 				"gpt": false,
> 				"mounted": false,
> 				"name": "sdd1",
> 				"size": 128320719872,
> 				"used": "filesystem",
> 				"uuid": "64A5-F009"
> 			}
> 		],
> 		"rpm": null,
> 		"serial": "0392523110004665",
> 		"size": 128320801792,
> 		"status": "unknown",
> 		"used": "partitions",
> 		"vendor": "Samsung",
> 		"wearout": null,
> 		"wwn": null
> 	}
> }
>
> note: this pen drive is brand new, got just unwrapped and passed through to my dev
> VM, and as such it's still coming with the formatting from factoring.
>
> Now, I first did not even expect that it shows up in the selector, but it did, so I'm
> wondering if it either should not be available or if it should work to use this disk
> too.
>
> No worries, I do not want an immediate fix or the like, just would like to know what's
> the expected outcome here is – as I think quite some other users might also plug in their
> freshly unwrapped and proudly exfat/vfat formatted pen drives to see how this goes.
> That they have to do something might be fine, but a regex not matching error won't
> shove them in the right direction I think.

Good point, we could either
 - not reutrn UUIDs that are not really UUIDs[1]
 - filter them out in the UI

In both cases the supported filesystem should probably be mentioned to the error msg.
Generally I think it'd be better to not return a UUID that is not really
a UUID. So since, for example, `exfat` only has a 'pseudo-UUID' it would
not be included in the returned data.

[1] https://www.rfc-editor.org/rfc/rfc9562.html#name-uuid-format