Qualified Functionality in ClusterStor 3.2
Functions supported or not supported in ClusterStor software version 3.2
- Support for Live Updates from ClusterStor 3.1 to 3.2
- The filesystem could remain online during the update
- Users should expect performance degradation during the update
- Batch update method (with the filesystem down) is supported
- OEM server updates are now supported
Note: Live updates from 3.0 to 3.2 are NOT supported. For these updates, the filesystem must be taken down. - The filesystem could remain online during the update
- Improved GOBI update script
- Reduced time to apply the firmware
- Automatic retries for failed actions (up to 3 times)
- Ability to control timeouts for each operation (transfer firmware, select firmware for activation, activate firmware) as well as timeout for retry operation. Cray does not recommend changing the timeouts below the default settings as it may affect stability of the update process. Default timeouts:
- transfer: 3 seconds
- select: 3 seconds
- activate: 60 seconds
- retry: 10 seconds
- Encrypted Data at Rest Key Management
- The format of the system's serial number changed to use Cray Asset Name
- For production systems the format of the serial number in the installation YAML file must match Cray Asset Name (e.g., 12345678)
- For non-production (lab) systems the format of the serial number remains the same (e.g., CSSD12345678)
- cscli csinfo output on production systems now reflects the format change:
cls12345n00: System Serial Number: 10011516 cls12345n00: Possible previous SSN: N/A cls12345n00: OEM System Serial Number: N/A cls12345n00: System Identifier: N/A cls12345n00: Cluster Name: cls12345n cls12345n00: Filesystem Name: cls12345 cls12345n00: Filesystem Type: Lustre cls12345n00: Hardware Platform: CS-L300 cls12345n00: Data Network Type: 40GbE cls12345n00: Software Release: 3.2 cls12345n00: Full System Update: N/A cls12345n00: RAS-only System Update: N/A cls12345n00: Firmware-only System Update: N/A cls12345n00: FS-only System Update: N/A
- Install the software release via the System Updater to an OEM server
- Install and deploy systems with SSU configuration and ESUs, with any combination of the following storage building blocks:
- SSU
- SSU+1, SSU+2, SSU+3
- L300F
- L300F SSU with GridRAID (hardware)
- This is a Storage Scalable Unit, not a standalone product
- High Availability:
- SSU node failover/failback
- SMU node failover/failback
- 2U MMU node failover/failback or optional 4U AMMU node failover/failback
- Dual Management Network (DMN) switch redundancy
- Lustre 2.11 with the following supported features:
- LockAhead (2.11)
- Progressive File Layouts (PFL) (Lustre 2.10)
- Project Quotas (Lustre 2.10)
- Large Bulk I/O (2.9)
- Subdirectory Mounts (Lustre 2.9)
- Multiple modify RPCs per-client (2.8)
- Online lfsck (Phase 4 with improved performance) (Lustre 2.8)
- O/S: CentOS 7.2 Linux base, plus CentOS 7.4 updates to address Spectre/Meltdown
- Support for sharing Brocade/Ruckus ICX6610 management switches with Cray clusters
- Local Data Network (LDN)
- InfiniBand EDR, FDR, QDR
- Support for Bonded Ethernet (active/passive, active/active )
- Omni-Path
- CSCLI – ClusterStor Command Line Interface
- ClusterStor Manager (CSM) browser-based GUI
- Supported browsers for Windows, Linux, and MacOS: Chrome, FireFox, Safari, Microsoft Edge, and Internet Explorer 11
- Support File Bundle Collection
- Lustre file system with two (2) storage tiers (flash and disk, L300F only)
- NXD I/O acceleration function (L300N only)
- RAID Stack: Updates to optimize GridRAID and SCSI performance
- Spectre and Meltdown security updates
- Support for GOBI OneStor USM
- Automated HDD and SSD firmware updates
- System Updater for providing basic rolling updates. Roll backs not supported. (Find more System Updater content in the ClusterStor™ Field Installation Guide H-6163 at http://pubs.cray.com.)
- Support for 3.2 on CS9000 and Sonexion 2000 GridRAID systems (CNG not supported), via CUP from 2.0 to 3.1 followed by a standard update to 3.2 from 3.1
- RAS features:
- RAS Infrastructure: CLI, Nagios and Ganglia plugins, REST API
- Guided Walkthrough Repairs: 2U24 / 4U24 / 5U84 Drives, 2U24 / 4U24 PCMs, 5U84 cooling module
- Fault Isolation: 5U84 cooling module, 2U24 / 5U84 I/O controllers
- Support for AMMU PSU failures
- Support for AMMU internal HDD
Hardware
- System Management Unit (SMU)
- Metadata Management Unit (MMU):
- MMU Addition procedure
- Optional Advanced Metadata Management Unit (AMMU) as a replacement for the base MMU:
- All-SSD AMMU available
- Support for up to two (2) AMMUs in a system
- AMMU Addition procedure
- Scalable Storage Unit (SSU):
- Support for two (2) or four (4) SSDs in each SSU
- SSU Addition procedure
- L300F:
- 2U24 enclosure with 24 SSDs
- L300F Addition procedure
- Embedded Application Controller (EAC) platform using Intel CPUs
- LDN interface:
- Mellanox CX-4 – EDR / FDR HCAs
- Mellanox CX-4 IB EDR/100GbE HCAs
- Intel Omni-Path 100 HCAs
- Disk drive support (dependent on installed drive model):
- Hard disk drive (HDD) capacities: 300GB up to 14TB
- Solid state drive (SSD) capacities: 100GB up to 3.2TB
- Drive capabilities:
- L300, L300N, Sonexion 3000, 3000T: 4K Native, T10-PI Format Type2, SED, and FIPS
- CS9000, Sonexion 2000: 4K Native, 512e, T10-PI Format Type2, SED, Non-SED, and FIPS
Features NOT Supported in ClusterStor 3.2
- Lustre features not supported:
- Multi-rail LNET (Lustre 2.10)
- Data on Metadata (DoM) (Lustre 2.11)
- File Level Redundancy (FLR) (Lustre 2.11)
- Dynamic LNet Config (Lustre 2.11)
- DNE phase 2 in Lustre is experimental and not enabled by default
- UID/GID Mapping (Lustre 2.9) is experimental
- CIFS/NFS Gateway (CNG)
- SELinux preliminary client support