the bucket on one cluster and wait for the bucket to be replicated to the other cluster. Failure to do
so results in objects contained in one bucket or the other being inaccessible.
• Named object deletes. Deleting a named object does not decrement the stream count displayed
at the top of the cluster status page on the Admin Console.
• Use curl version 7.19.7 or later. If you use curl with DX Storage, and you use the new
authorization feature, you must use curl version 7.19.7. There are known issues with earlier
curl versions.
• Last-Modified Header with Previous DX Storage Versions The Last-Modified header is not
currently returned for streams written with DX Storage versions earlier than 4.0. New streams and
anchor stream updates written with 4.0 or later do contain the header.
• Vol Number on Admin Console The Vol # labels next to each DX Storage node in the Admin
Console are listed in arbitrary order. The Admin Console labels do not correspond to physical
drive slots in the machine chassis. However, the volume names do match the physical drives in
the machine chassis.
• Consumer-Grade Disk Drives. Some non-enterprise-class disk drives have lengthy error
recovery logic. When an error occurs on these types of drives, it might take minutes for a disk
read or write operation to complete. In these cases, the client could see very long response times,
or it might even see a socket timeout if the delay is too long.
Many enterprise or server grade disks are designed to return errors within a limited period of time,
allowing recovery or rebuild operations to begin immediately and to eliminate the lengthy delays
on I/O operations. Dell recommends against the use of consumer grade disk drives and does not
support them in a high demand environment.
• Avoiding Client Timeouts with Large Streams. Client operations with large streams (1GB or
larger) can take several seconds or more, depending on stream size. Clients that support large
stream operations should set their socket timeouts accordingly to avoid client timeout errors.
• Network Time Protocol. Dell strongly recommends you identify an NTP time source in at
least one of your cluster or node configuration file, as discussed in the DX Object Storage
Getting Started Guide. Without NTP, the individual clocks on the nodes of the cluster will still
be synchronized to each other, but they will likely drift over time so that the cluster will have
the wrong time. This will impact stream storage policies such as retention and replication as
discussed in the following bullet.
• Time Clock Synchronization. When formatting storage policies in lifepoint headers, it is very
important that the local clock be reasonably accurate so the end dates of the lifepoints reflect
the true UTC time. The DX Storage cluster itself can (and should) synchronize itself to an
accurate time source as specified in the DX Storage Getting Started Guide. If the client mistakenly
specifies the wrong end date in a stream's storage policy, perhaps because its local clock is set
incorrectly, there could be unintended consequences, including premature deletion, when DX
Storage enforces that policy.
• Replica Terminology. The term replica has special meaning in the context of DX Storage. All
instances of a stream stored within a DX Storage cluster are identical – there is no original.
Therefore, saying there are two replicas of a stream means there are exactly two identical
instances (not an original plus two copies).
• Not Found Errors in a Busy Cluster. A READ, INFO, UPDATE, or DELETE request to a heavily
loaded cluster might rarely result in a 404 Not Found response, even if the requested object is
Copyright © 2010 Caringo, Inc.
All rights reserved
4
Version 5.0
December 2010
ページを印刷する