QUALITY OF SERVICE
SMB
S3
PLATFORM & CONTROL
LICENSING
S3 Connection Limit in User QoS Policies
VAST Cluster 5.2.1 lets you restrict the maximum number of S3 connections to the VAST cluster that can be opened by a client user. This helps prevent scenarios where a single client user creates an enormous amount of S3 connections exhausting cluster's TCP connection resources.
This feature is enabled by default (but no limit is applied unless you configure the limit in a user QoS policy). To disable it on your VAST cluster, contact VAST Support.
To set an S3 connection limit in a user QoS policy:
Enter the maximum allowed number of S3 connections in the S3 Maximum Connections field in user QoS policy settings (Element Store -> QoS Policies -> choose to create or edit a user QoS policy -> User tab).
Run the
qospolicy createandqospolicy modifycommands with the--s3-connections-limitoption specified.
The following limitation applies:
(RESOLVED IN 5.2.2) ORION-250317: In case of a CNode HA event, the cluster needs a few minutes to retrieve connections from the failed CNode. During this time, the user will be served at the user's S3 connection limit less the number of connections that were provided by the failed CNode. For example, if a user had a limit of 100 connections, 10 of which were served by CNode A and the remaining 90 were on other CNodes, right after an HA event of CNode A the user will only have 90 connections. After a few minutes, the number of available connections will be back to 100.
For more information about this feature, see VAST Cluster Administrator's Guide.
Support of SMB Hardlinks
VAST Cluster 5.2.1 introduces support for creating and managing hardlinks on SMB shares exposed by the VAST cluster. (Prior to this change, an attempt to create a hardlink on VAST SMB share would be rejected with the STATUS_NOT_SUPPORTED error.)
You can perform basic operations on the hardlinked files, e.g. read, write, set ACLs, copy, move, overwrite, unlink and delete. Nested hardlinks (hardlink to a hardlink or a softlink) are also supported.
The following limitations apply:
Creating a hardlink to another SMB-enabled view is not allowed.
Creating hardlinks to directories is not allowed. An attempt to link a directory returns a STATUS_FILE_IS_A_DIRECTORY error.
Linking an Alternate Data Stream (ADS) is not supported and returns a STATUS_INVALID_PARAMETER error.
Creating a hardlink on a Global Access satellite cluster is not supported and returns a STATUS_NOT_SUPPORTED error.
Creating a hardlink, on the Global Access origin cluster, of a file that has already been opened from a satellite cluster, is not allowed and returns STATUS_INTERNAL_ERROR.
By default, both creating and opening of SMB hardlinks are enabled on the cluster. If you want to configure the cluster so that it does not allow creating and/or opening SMB hardlinks, contact VAST Support.
S3 Bucket Monitoring
VAST Cluster 5.2.1 provides metrics to help monitor and analyze S3 access to buckets exposed by the VAST cluster. The metrics include:
Per tenant: bandwidth (number of bytes uploaded/downloaded) and IOPS (total number of requests over time)
Per bucket (view):
Total number of requests responded with a 4xx or 5xx HTTP error code for the client and the server
Total amount of requests per request type (PUT, GET, DELETE, POST, HEAD)
The metrics are also provided for replication and/or Global Namespace destination buckets.
Note
In case of a CNode HA event, the metrics are reset.
The metrics can be queried via the VAST REST API.
The following limitation applies:
ORION-240888: S3 bucket monitoring does not take into account VMS-originated S3 requests, including those related to features such as:
Lifecycle rules
Bucket notifications
Bucket policies
Object ownership
Bucket versioning
Object locking
Bucket logging
Known issues include:
(RESOLVED IN 5.2.2) ORION-242283: An invalid ACL is not counted as a client error in the S3 bucket metrics.
For more information about S3 bucket metrics, see VAST Cluster Administrator's Guide.
Rack-Level Resiliency
VAST Cluster 5.2.1 offers rack-level resiliency. When a cluster is configured to be rack-level resilient, it is capable of maintaining full data availability even when an entire rack of DBoxes goes down.
To implement rack-level resiliency, deploy the cluster using the VAST CLI cluster create command with the --rack-config option specified. This option lets you supply a JSON configuration file that states which DBoxes are available in which rack.
The racks defined in the JSON configuration file are considered failure domains. If one of the failure domains fails, the cluster will continue to provide service without disruption.
To verify the rack-level resiliency status on a running cluster, open cluster settings in VAST Web UI (Settings -> Cluster -> Data Management) and check if the Rack level resiliency option is enabled.
The following rules and limitations apply:
Every DBox must be associated with a single failure domain.
At least seven failure domains must be defined.
The total number of DBoxes in each failure domain must be the same +/- 1.
If two of the racks in a rack level-resilient cluster go down, bringing up only one of them does not return the cluster to normal operation. The cluster restores service only after the second of the racks goes up.
Note
Implementing rack-level resiliency automatically enables the DBox HA feature for the cluster.
For more information about this feature, see VAST Cluster Administrator's Guide.
CPU Core-Based Licensing
With release 5.2.1, VAST introduces licensing based on the number of CPU cores managed by the cluster. These include CNode and EBox cores, as well as external engine compute resources that will be managed by VAST in the future.
The number of cores covered by the existing license is displayed in the Support -> License page of VAST Web UI together with other license details, as well as in the output of VAST CLI license list and license show commands.