Problem Note 10468: Possible incorrect results when using &SPDSEV1T and &SPDSEV2T
Setting &SPDSEV2T=2 forces SPD Server to use parallel reads when
returning data even if the data is sorted. This option does not mask
the sort order from the SAS client. This can lead to situations where
the sort order is expected by the client application, but records are
not returned in sorted order. Errors and incorrect results can be
produced. SQL Joins and GROUP BY clauses have been identified as a
possible processes affected by this.
Note that as documented in the SPD Server documentation, "this problem
does not apply to SQL statements executed via the SPD Server
pass-through SQL mechansim".
As a result, we recommend not using &SPDSEV2T=2. If you have sorted
data and want to have the SPD Server return data in sorted order, use
the dataset option SORTEDBY=_NULL_. This will hide the sort order from
both the client procedure and SPD Server, freeing it up to use parallel
reads.
SPDSEV1T controls whether parallel reads are used when reading indexed
data. Using &SPDSEV1T=2 is discouraged for similar reasons. Instead,
we recommend using &SPDSNIDX=YES to have the server block index usage.
Operating System and Release Information
| Product Family | Product | System | Reported Release | Fixed Release* |
| SPDS | Scalable Performance Data Server | Solaris | 3.0 | |
| Microsoft Windows NT Workstation | 3.0 | |
| 64-bit Enabled Solaris | 3.0 | |
| HP-UX | 3.0 | |
| 64-bit Enabled HP-UX | 3.0 | |
| Tru64 UNIX | 3.0 | |
| 64-bit Enabled AIX | 3.0 | |
| AIX | 3.0 | |
*
For software releases that are not yet generally available, the Fixed
Release is the software release in which the problem is planned to be
fixed.
| Type: | Problem Note |
| Priority: | high |
| Date Modified: | 2003-11-26 11:07:28 |
| Date Created: | 2003-07-21 11:46:17 |