CSN Platform Server Release Notes

The Swarm CSN Platform Server handles installation and configuration of both the network services required to run a Swarm cluster and the Swarm software used to interface with it into a single integrated server. 

Required Actions

CentOS 6 Repo

CentOS 6 expired on November 30, 2020, so the repository references on the CSN must be updated to point to the vault. The file is located here:

  • /etc/yum.repos.d/CentOS-Base.repo

To fix the references, download the latest Support tools (swarm-support-tools.tgz), which includes the corrected CentOS-Base-6.repo file, and put it in the CSN’s /etc/yum.repos.d folder.

Reformat dnsServer

The dnsServer setting uses the legacy dictionary format, which needs to be trimmed off, if a single network CSN is installed:

dnsServers = ['10.1.1.6'] # Legacy format dnsServers = 10.1.1.6 # Correct format

The legacy format is unsupported and fails silently. This can cause clock skew if using a FQDN for the timeSource. (CSN-2118)

Backup/Restore

The manifest UUID must be present in the cluster in order for a CSN Restore action to work. Take these steps to verify the manifest UUID exists in the cluster before using the backup/restore feature. (CSN-2117)

  1. Retrieve the manifestuuid from the bootstrap file, which only contains the UUID: 

    /var/opt/caringo/csn/backup/manifest/.bootstrap
  2. Check whether the manifest UUID is present in the cluster. Run this command:

    curl -IL [nodeip]/[manifestuuid]
  3. Regenerate the manifest UUID if it is not present in the cluster:

    1. Remove the UUID from .bootstrap file.

    2. Return to the CSN UI.

    3. Create a new backup, which generates a new manifest UUID.

  4. Create a manual backup to verify it works.

Changes by Release

Release 8.3.2

The 8.3.2 package includes a fix for a security hole that affects RHEL/CentOS 6.10. (CSN-2113)

  • Upgrade impact: The MIBs on the CSN must be updated before upgrading to Storage 10.x. Run the platform-update-mibs.sh script from the Swarm Support tools bundle. (CSN-1872)

Release 8.3.1

The 8.3.1 package supports upgrading and installing on RHEL/CentOS 6.8 or newer. (CSN-2107, CSN-2112)

Prime the aliases for the metrics indices if upgrading the CSN 8.3.0 Swarm UI to RHEL/CentOS 6.9 or newer before installing CSN 8.3.1: 

/usr/share/caringo-elasticsearch-metrics/bin/metrics_curator -n -v
  • Upgrade impact: Customizations made to castor_net in the kernel options of the netboot.cfg are lost on upgrade and must be restored; other changes (gateway, maxProcesses) may not be preserved. (CSN-2081)

Release 8.3

Subclusters

As of Swarm Storage 9.3.1, 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. (SWAR-7675)


New Swarm Storage UI: CSN now includes the redesigned storage interface so the familiar CSN software stack can be used while transitioning from the legacy Swarm Admin Console.

  • The new UI is available from the URL http://{CSN·host}:91/_admin/storage and includes the ability to view, configure, and manage the storage cluster. (CSN-2057)

  • When deployed on a dual-network CSN, a new Service Gateway component bridges the communications between the front-side network and the back-end network that connects the storage nodes. See Installing the Storage UI.

  • The legacy Swarm Admin Console is still available during the transition to the new UI.

  • See Swarm Storage UI in the Swarm Storage guide.

Historical Metrics hosted on CSN: CSN includes all components needed to allow the CSN to collect and store the operational time-series data produced by the storage cluster and used by the new UI. This means that deployments that do not make use of a separate Elasticsearch cluster for metadata searching can now collect these metrics.

  • The software components and configuration to host a metrics-only Elasticsearch database are included in the CSN software bundle. This is intended for deployments that do not already have a separate Elasticsearch cluster for metadata searching. (CSN-2060)

  • After starting historical metrics collection, it can take a minimum of 30 minutes before the new UI can render trend graphs based on the data.

  • See Swarm Historical Metrics in the Swarm Storage guide.

Other changes:

  • The CSN Console (http://{CSN·host}:8090) has removed the legacy Reporter functionality, which is replaced by Swarm Metrics in the Swarm Storage UI. (CSN-2058)

  • The CSN Console has removed the options for persisted settings, which are no longer configurable from the CSN Console but are settable in the Swarm Storage UI.

  • CSN includes updated NTP packages. (CSN-2016)

Release 8.2

  • CSN has been updated to install on RHEL/Centos 6.8.

  • The CSN software bundle includes a version update to Swarm 8.2.1. Replication configuration settings are automatically migrated to work with new Swarm domain and bucket-level policies as part of the CSN upgrade process. See Configuring Cluster Policies.

  • The CSN software bundle includes a version update to SCSP Proxy 8.2. 

Release 8.1

  • The CSN software bundle includes a version update to Swarm 8.1. Erasure coding configuration settings are automatically migrated to work with new Swarm domain and bucket-level policies as part of the CSN upgrade process. See Configuring Cluster Policies.

  • Due to breaking changes in the cluster.cfg EC settings for Swarm 8.1, CSN 8.1 and higher prevent restoring CSN backups from versions prior to 8.1. (CSN-2025)

Release 8.0

  • The CSN software bundle includes a version update to Swarm 8.0. Content Router is no longer included or supported.

  • An error in the CSN's support data collection tool, sosreport, has been corrected to ensure all the necessary information is available when filing a support ticket. 

Limitations and Known Issues

These are the known issues and operational limitations that exist in this release of Swarm CSN.

  • Invalid config parameters that prevent the unassigned nodes from booting are created subcluster assignments are removed in the CSN UI. (SWAR-7675)

  • If you install a single-network CSN on a machine that has multiple NICs that are live, it fails silently. (CSN-2116)

  • Changes made to /etc/caringo/netboot/netboot.cfg for the default line of "kernelOptions = castor_net=balance-alb:" are not preserved through upgrades to CSN. (CSN-2081)

  • Upgrading to CentOS 6.8 while the EPEL repo is enabled upgrades monit to version 5.14.1, which breaks the firstcsnboot script. Do not enable the EPEL repo until after the CSN has been installed. (CSN-2053)

Upgrading the CSN

Important

Configuration backups taken prior to version 8.2 are not compatible with version 8.2 due to both the end of life of Content Router and settings changes required in Swarm 8.2. 

Upgrade to RHEL or CentOS 6.8 to upgrade a version 3.x or later CSN and then install the new software using the included installation script. yum updates need to be re-enabled before upgrading.

See CSN Upgrades.

Note

CSN 7.0.1 was the last release for which upgrade from 2.x was supported. To upgrade from 2.x requires two steps: upgrading first to 7.0.1 and then again to the latest version. 

Application and Configuration Notes 

These CSN configuration issues require special attention:

  • SNMP errors related to _ifXTable_container_row_restore. In some cases errors similar to the following might display when installing CSN: error finding row index in _ifXTable_container_row_restore. This error is harmless and can be ignored. See Red Hat bug 736264.

  • Messages to Ignore during Installation and Initial Configuration. During the NIC portion of initial configuration, messages similar to 'eth0 NIC Copper Link is Down' may display. Service Fail messages might display before CSN reboots after CSN is installed and initially configured. Additionally, there may be logged errors related to an inability to resolve FQDN names. These messages are all harmless having to do with either interim system status as the configuration is applied or timing related states that are retried. The messages can safely be ignored.

  • Clean System Start State. The CSN assumes a clean system start state prior to installation, with no previous network configuration of NICs and ports. In particular, manual pre-configuration of NIC aliases (ifcfg-eth1:1) may lead to unexpected network behavior on the CSN as the network configuration performed during installation are not aware of the pre-configured aliases.

  • RedHat Administrative Tool Compatibility. Both the RedHat Network Administration Tool and the RedHat Security Level Configuration Tool are incompatible with the CSN because they alter the expected location and/or configuration of critical network and firewall configuration files, possibly interfering with initial CSN configuration or communications. To prevent conflicts, these tools are removed during installation of the CSN.

  • SELinux Errors. While SELinux is run in permissive mode during the bootstrap process, SELinux errors may still appear during bootstrap. These errors can be safely ignored and do not occur again after a system reboot.

  • Monit Configuration. The monit watchdog process is carefully configured on the CSN to ensure compatibility with different products and services. Manual configuration of monit parameters is not supported.

  • SSH Delays Logging into CSN. Administrators may see delays logging into the CSN via SSH if there are any non-existent or unreachable DNS servers specified in the resolv.conf file. Adding a "UseDNS no" line to the /etc/ssh/sshd_config should remove the delay.

 

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