This facilitates upgrading your cluster without scheduling an outage or bringing down the cluster
when needed. Just restart your nodes one at a time with the new version and the cluster will
continue serving applications during the upgrade process. If you have stored your streams with at
least two replicas, they will continue to be fully accessible during the upgrade.
For a simpler upgrade strategy, the entire cluster can be rebooted at once after the software on
the centralized configuration location has been updated. Please see the Upgrading appendix
of the DX Object Storage Administration Guide for additional information on these upgrade
strategies.
Note
Dell strongly recommends that clusters with DX Storage volumes originally formatted with
releases earlier than version 2.0 (September 2007) be upgraded to the 5.0 release as
soon as possible so the volume recovery process from recent software versions remains
compatible with the older disk format.
1.4. Limitations of this Release
These are the known issues and operational limitations that exist in this release of DX Storage.
• DX Storage can create too many replicas of large objects on empty volumes. Testing shows this
issue affects streams of 50MB or larger. Using the Admin Console's Cluster Settings page to
increase the value of Multicast % to 10% or more helps minimize the issue. This issue is expected
to be resolved in a future release.
• Before you rename a domain using the Admin Console, make sure there is no domain with that
name already.
• There is a name size limit of 8192 characters, after percent encoding, for a bucket or named
object.
• DX Storage does no syntax validation on Castor-Authorization header syntax at the time a
new object is stored.
• There is a security vulnerability that could be exploited to gain read access to streams in cases
where read access is otherwise prohibited by the stream's content authentication policy. The
vulnerability provides no risk for data corruption. Exploiting the vulnerability requires prior
knowledge of the stream UUID or name.
• A GET that follows a GET with a ?newhastype query argument occasionally fails.
• A node that loses connectivity with its syslog host for an extended period of time might log
invalid argument traceback errors to the DX Storage administration console. Administrators
who see these errors should check connectivity to the configured syslog host.
1.5. Application and Configuration Notes
Special attention should be paid to the following items when developing DX Storage applications or
configuring DX Storage clusters. More details can be found in the DX Object Storage Application
Guide and the DX Object Storage Advanced Administration Guide.
• Duplicate bucket creation in mirrored clusters. In a certain cluster configuration referred to as
active-active, do not create the same bucket in the same domain in each cluster. Instead, create
Copyright © 2010 Caringo, Inc.
All rights reserved
3
Version 5.0
December 2010
Imprimer la page