Changes for Storage 9.6

Owned by Rumena Zlatkova (Deactivated)
Last updated: Apr 06, 2022 by Kyle Miller (Deactivated)
4 min read

New Features

This release offers the next generation of replication technology, extensive OS software updates, and a broad collection of performance, error-handling, and consistency improvements across Swarm.

Replication via Direct POST

Swarm Storage now supports remote replication via direct POST, so replication feeds can operate directly through Content Gateway 5.3. When a replication feed is defined, specify which replication mode to use: either the legacy bidirectional GET method of replication (which may be needed for specific application compatibility or network requirements) or the new direct POST method, which offers better performance and flow management. The Swarm Storage UI supports the new direct POST method of replication. Each replication feed shows which type of replication is in use and what speed (number of concurrent threads) is set for direct POST:

See Managing Feeds. (SWAR-7804)

OS Software Updates

For this release, the Linux kernel, Linux firmware, and numerous driver versions are updated, in addition to network drivers and additional kernel modules to support modern hardware.

  • Intel network drivers are updated to support testing with newer Ethernet adapters.

    • i40e to 2.4.10 and ixgbe to 5.3.7. (9.6.4: SWAR-8151)

    • i40e to 2.4.3 and i40evf to 3.4.2. (9.6.3: SWAR-7643)

  • Additional kernel modules (edac_core, iosf_mbi, and shpchp) are now included to address PCIx address mapping issues with newer Lenovo mass storage devices. (SWAR-7937)

  • Updated versions (SWAR-7935):

    • Linux Kernel: 4.14.34

      • Note: Swarm does not include or support the Liquidio and Netronome families of network drivers. (SWAR-8037)

    • Linux Firmware: 1.173

    • Intel ixgbe driver: 5.1.0-k

    • Intel igb driver: 5.4.0-k

    • Intel e1000e driver: 3.2.6-k

    • Intel ixgbevf driver: 4.1.0-k

    • Intel i40e driver: 2.1.14-k

    • Intel i40evf driver: 3.0.0-k

    • hpsa driver: 3.4.20-0

    • megaraid_sas driver: 07.702.06.00-rc1

    • mpt2sas driver: 15.100.00.00

    • mpt3sas driver: 15.100.00.00

  • Added support for HPE ProLiant Gen10 P-series Smart Array controllers (SmartPQI). (9.6.3: SWAR-8128)

Performance and Error Handling

  • This release includes reductions in both the administrative network traffic between the Swarm cluster and Elasticsearch nodes and the ERROR-level logging for this traffic. (SWAR-7470)

  • Boot times are shortened by better entropy generation during startup. (SWAR-7598)

  • Swarm now has improved and proactive handling of search feeds that are blocked by hidden causes, such as stuck plugins. (SWAR-7890)

  • Swarm Storage has numerous improvements to memory management and its ability to detect and diagnose those problems. (SWAR-7925)

Consistency and Clarity

  • Rebooting and shutting down a chassis now establishes for the cluster that the affected volumes are in maintenance mode and so are inaccessible during the down time. (SWAR-7964)

  • For consistency across all types of usage reporting, the formulas for calculating percentage of cluster capacity used are standardized across the Storage UI, the Admin Console cluster status page, and the cluster health (phone home) report. (SWAR-7959)

  • When using Swarm's serial port console to format volumes for use in the cluster, the formatting command now reports the success and suggests next steps. (SWAR-6494)

  • For parity with replication feed definition in the Swarm UI, the legacy Admin Console (port 90) now includes a field for specifying a custom port to be used for the target cluster. (SWAR-7994)

Upgrade Impacts

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

Impacts for 9.6

  • Upgrade required for 9.6.0 and 9.6.1 — Upgrade to fix a potential EC issue that can affect long-term data protection (9.6.2: SWAR-8111). 

  • Upgrading from 9.2.0 or earlier — Because Swarm no longer generates the deprecated header Expect: Content-MD5, rolling upgrades in and remote replication with clusters running 9.2.0 or earlier are not supported. Recommended: Upgrade to 9.3.2 first, then upgrade to the current version. (SWAR-7742)

  • EC and rolling upgrades — Because of changes to EC manifests in 9.4, sites performing EC writes by APPEND and Multipart APPEND (unlikely and rare) should contact DataCore Support before performing a rolling upgrade from a version older than 9.4. Note, too, that the HP may report errors replicating EC manifests during the upgrade, but data integrity is not affected, and the errors stop when the upgrade is complete. (SWAR-7767)

  • SwarmFS Configuration Changes — If the storage cluster parameter scsp.keepAliveInterval=45 is set to use SwarmFS with Storage 9.5.0, restored the 15 second default. (SWAR-7917)

  • Content Gateway 5.2.3+ required — If using Content Gateway, upgrade to version 5.2.3 or greater immediately after upgrading Swarm Storage.

  • Swarm Search and Metrics 2.5+ required — Swarm 9.5 and higher requires an upgrade of Swarm Search and Swarm Metrics, both of which need to be version 2.5+. (SWAR-7818)

Additional Changes

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

  • Resolved Issues

    • For uncompleted parts of a multipart completion request, HP may log erroneous "DeletionConstraint / TypeError" error messages. They ceased when the request completed, but incomplete operations that timed out required deleting the uploaded parts to eliminate the error messages. (9.6.1: SWAR-8082)

    • A COPY on an erasure-coded object fails if requesting a Content-MD5. This affects the metadata editing feature introduced in Content UI 5.4. (SWAR-7980)

    • Some rapid updates can lead to a blocked Search feed. (SWAR-7969)

    • In replication feed definitions, the regular expression special character '+' cannot be used in domain or domain exclusion patterns. (SWAR-7873)

    • Clusters with retired volumes initiate volume recovery during a rolling reboot. (SWAR-7968)

    • The response for a bucket rename operation returned the wrong domain query argument on the Location header. (SWAR-7943)

    • Long-running Swarm processes can cause SNMPSTAT_RESPONSEHISTOGRAM_TOTAL_SAMPLES errors to be logged. (SWAR-7938)

    • Using a metadata value like NAN, INF, or INFINITY (regardless of case or -/+ prefix) caused a FEEDS ERROR with a mapper_parsing_exception and prevented the object from being returned in listings. (SWAR-7913)

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