Changes for Storage 9.3

New Features

This release has several enhancements to improve the operation and ease of maintenance of the storage cluster.

  • Filtered replicationSwarm provides finer granularity and greater control when replicating content between clusters. Replication feeds support multiple filters to specify which domains to include, which ones to exclude, or both, and provide individual controls for replicating unnamed objects with no domain. The domain filters support wildcard matching for uncomplicated rule specifications. (SWAR-7398)


    See Managing Feeds.

  • Dynamic Stale Volume Handling Administrators have an easier mechanism for handling stale (offline for over two weeks) volumes without rebooting the cluster. Stale volumes can be dynamically mounted or reformatted while retaining the volume's encryption status. (SWAR-6758)
    See Returning a Stale Volume to Service.

  • Faster Random Reads of Large Objects - Range read performance of large files is improved using caching of erasure coded segments. (SWAR-7570)

Upgrade Impacts

These items are changes to the product function that may require operational or development changes for integrated applications.

Impacts for 9.3

  • Legacy authentication turned off by default If the integration requires Swarm storage’s legacy auth/auth mechanism, setting "security.noauth = False" in the cluster configuration is necessary before booting this new software version. DataCore recommends that customers migrate to Content Gateway’s more comprehensive access control system. (SWAR-7399, SWAR-7115)

  • Content Gateway 5.2.1+ required - For deployments that use the Content Gateway, upgrade to Content Gateway version 5.2.1 or greater at the same time the storage cluster is upgrading.

  • CSMeter incompatibility - Because of requirements for Elasticsearch 2.3.3, the legacy CSMeter software component of CloudScaler 4.0 and earlier is incompatible with Swarm 9.0 and higher. The improved metering and quota functionality built into Content Gateway has replaced CSMeter.

  • Subclusters

    • Default subcluster - The default subcluster name has changed to "default" rather than blank. Note: if subclustering is not used, while multiple subclusters may appear during the upgrade, there are no post-upgrade actions required. (SWAR-7526)

    • CSN - Subcluster assignments can no longer be blank, and CSN installations with mixed subcluster assignments have the unassigned nodes unable to boot, showing an error in contacting the time source. Supply a subcluster for each node if any named subcluster is specified in the cluster.

  • Disabled health reports - If cluster health reports is disabled by setting support.uri to blank, Swarm fails to boot. Delete the setting entirely and set support.reportPeriod=0 to suppress reporting without errors.(SWAR-7644)

  • scsp.maxParallelGetRetrieveTasks - The setting scsp.maxParallelGetRetrieveTasks is no longer public and has been removed from the Swarm UI. If scsp.maxParallelGetRetrieveTasks is customized, contact DataCore Support and use snmpset or curl to change it. (SWAR-7560)

  • scsp.autoRecursiveDelete - scsp.autoRecursiveDelete now defaults to True, automatically reclaiming the disk storage for child content of buckets and domains upon deletion of the parent context. (SWAR-7551)

  • createdomain - The createdomain query argument is deprecated and replaced by the Content-type: application/castorcontext header used by the Content Gateway. See Manually Creating and Renaming Domains.

Additional Changes

These items are other changes and improvements including those that come from testing and user feedback.

  • OSS Updates

    • none

  • Improvements

    • The volume of INFO-level log messages for certain internal tasks has been reduced. (SWAR-7595)

    • The descriptions for Swarm Storage configuration settings now include examples where needed, such as for how to format multiple values. (SWAR-7536)

    • Swarm documentation references and links to the latest HTTP 1.1 specification sections (RFC 723x) for HTTP methods, headers, and response codes. (SWAR-7393)

  • Resolved Issues

    • An issue with the Intel 'igb' network driver led to non-optimal IRQ balancing in non-NUMA CPU architectures. (9.3.2: SWAR-7163)

    • Formatting volumes via the server console now allows Swarm processes to start without a reboot of the chassis. (SWAR-7544)

    • Incomplete multipart upload parts are not reclaimed when the domain is deleted recursively. (SWAR-7563)

    • Deletion of a misconfigured search feed took a long time to complete, leaving the feed visible after the operation should have succeeded. (SWAR-7307)

    • DELETE requests on very full nodes/clusters returned a 500 Internal Server Error response instead of 503 Service Unavailable. (SWAR-7569)

    • Multipart complete by PUT or APPEND requests succeeded even when the basis object had been deleted. (SWAR-7504)

    • Policy-*-Evaluated-Constrained headers are sometimes duplicated in SCSP GET and HEAD requests on domains and buckets. (SWAR-7585)

    • Rapid updates of mutable objects in clusters with the overlay index sometimes returned 404 Not Found errors. (SWAR-7568)

    • With multipath enabled, rapid removal of multiple volumes can result in some appearing as retired when added back. (SWAR-7552)

    • The Health Processor does not scan an entire volume and the console shows the error "'Accounting Cursor' object has not attribute '_attachTopage'". (SWAR-7557)

    • When the total number of indexed items on a node was near the capacity of the overlay index, Swarm erroneously returned 404 Not Found on some requests or 503 Service Unavailable on writes due to WriterOutOfMemory. (SWAR-7539)

    • During rolling upgrade and rolling reboot involving 9.0.x releases, critical log messages reported attempts to contact rebooting nodes. (SWAR-7317)

    • In nearly full clusters, client writes sometimes failed with a 507 Insufficient Storage response code. (SWAR-7324)

© DataCore Software Corporation. · https://www.datacore.com · All rights reserved.