Netezza Performance Server 11.2.1.13 release notes
Version 11.2.1.13 introduces enhanced recovery for SPU unresponsiveness and other stability fixes, aiming to provide a more efficient, reliable, and user-friendly experience.
New features and enhancements
- Improved performance for host-intensive queries
-
Increased the value of
host.snHostMemoryQuota
to enhance performance for host-intensive queries. - Enhanced recovery mechanism for SPU unresponsiveness
-
A faster recovery mechanism has been added to handle situations where the SPU becomes unresponsive due to a kernel dump. This feature includes a new registry setting,
enableKernelCore
, which allows for the kernel crash to be dumped along with the dmesg. The default value for this setting is "yes". - Configurable return size for to_char function
-
A new variable,
WAG_SIZE
, has been introduced to set the return size of theto_char
function. By setting a lower value for this variable, users can avoid the error "Record size 72913 exceeds internal limit of 65535 bytes". - Retry logic for database name retrieval
-
A retry logic has been added to the database name retrieval routine to prevent
preparePgCatalog
from failing if Netezza Performance Server is unable to connect to PostgreSQL in standalone mode. The routine will attempt to connect up to 3 times before failing.
Resolved issues
-
Memory leak: Resolved a host out of memory issue caused by a bit vector memory leak.
-
SPU oversubscription: Fixed queries failing with "SPU oversubscribed memory" error due to increased anonymous memory.
-
Issue of drives remain in locked state: Drives no longer remain in a locked state after drive replacement or activation of failed SPUs.
-
Postgres crash issue: Resolved the issue of
postgres
crash occurred when receiving aSIGQUIT
signal while running a UDF query. -
SSL query performance: Improved load and unload query execution time to match non-SSL query performance.
-
StatusUpdate.json corruption: Fixed corruption of
statusUpdate.json
due to concurrent access by parallel backup processes. -
NRS ETL Job Failure: Resolved NRS ETL job failures after role reversal due to insufficient permissions during restore.
-
NPS stuck in Pausing state: Fixed NPS getting stuck in the "pausing" state and implemented faster recovery when SPUs become unresponsive.
-
View restore failure: Resolved view restore failures caused by rewritten view definitions.
-
nzlogcollector error: Removed deprecated
nzpush -a mpath -issues
call fromnzlogcollector
on CPDS systems. -
Backup failure: Fixed backup failures due to objects being created in restricted schemas, starting from version 11.2.1.13.
-
NPS hang issue: Resolved an NPS hang issue caused by an extended
spuDumpTimeout
during an SPU core dump. -
Page table corruption: Increased the values of registry setting variables
rsSpuWriteFlushReserveMB
andrsSpuWriteFlushKBPerSlice
to improve memory allocation and addressing, preventing errors caused by inconsistencies or corruption in page tables, and providing a more robust system to handle high memory pressure. -
External table error: Fixed "
processExternalScanDataReader
: Invalid argument" error when loading from external tables. -
Cross-Database UPDATE: Fixed incorrect
UPDATE 0
response when performing cross-database updates. -
Sysmgr crash: Prevented
sysmgr
crashes duringnzstart
when zero rows are returned forsysmgr
catalog tables. -
Reduced restart time: Reduced restart time after postmaster crash by optimizing the find core process.
-
/tmp getting full: An issue causing the /tmp directory to become full due to multiple running
nzlogcollector
processes has been resolved.