HP Integrity NonStop H-Series Guida Utente Pagina 244

  • Scaricare
  • Aggiungi ai miei manuali
  • Stampa
  • Pagina
    / 346
  • Indice
  • SEGNALIBRI
  • Valutato. / 5. Basato su recensioni clienti
Vedere la pagina 243
HP NonStop Shadowbase Command Definitions
Database Specification Command Descriptions
HP NonStop Shadowbase Command Reference Manual—785421-002
235
NOSTOPONMISSING or STOPONMISSING – If the CONS detects that
blocks (records) are missing from a target during replication,
STOPONMISSING will result in the CONS abending with an
appropriate EMS message to indicated the condition. It will not abend
if NOSTOPONMISSING is configured.
BUFFERED or WRITETHRUCACHE – WRITETHRUCACHE causes the
CONS to set this mode of I/O for the file in the file system. This results
in blocks being flushed to disk immediately after a CONS does an I/O
for a given block. BUFFERED will cause the file system to use the
disk cache and flush less frequently.
Notes: Because I/O is not done to the alternate key files, you will
need to load the alternate key files before making use of the target
files in an application program that uses the alternate key files.
Because unaudited/unstructured block-oriented I/O is used in this
mode of replication, there is a possibility that some data loss can
occur if the NonStop system crashes and it becomes necessary to
deploy a target on a backup NonStop system. The most reliable
configuration to use in order to minimize this data loss is to set
NODELAYFLUSH and WRITETHRUCACHE. Note that CONS I/O
throughput is less with these settings. The best performing
configuration is to set DELAYFLUSH and BUFFERED, but this has
the highest probability of data loss should a failover to a backup
system become necessary.
For partitioned files, there is a window of time where data may flow
into a subsequent partition before the prior partition is full. If a HP
Shadowbase failure occurs before all data arrives for the prior
partition and you need to switch to a backup system application,
there may be some data loss. Additionally, an application will not
be able to sequentially read the data in the secondary partition
unless the prior partition is repaired. The FIXESEQ program is
included that will repair incomplete partitions. It should be run prior
to deployment of the target file being used in an application. For
more information about FIXESEQ, reference the HP NonStop
Shadowbase Installation and Planning Manual section on this topic.
After running FIXESEQ and before deployment of a target in an
application, you should turn audit on using FUP. You should also
consider doing a TMFCOM DUMP FILES of the related entry-
sequenced files after audit is turned on and before deployment of
the files. This is a consideration because the I/O used by the
Vedere la pagina 243
1 2 ... 239 240 241 242 243 244 245 246 247 248 249 ... 345 346

Commenti su questo manuale

Nessun commento