NOSORTINDEX 0/1 session

оНГБНКЪЕР БШЙКЧВХРЭ АШЯРПНЕ ОНЯРПНЕМХЕ ХМДЕЙЯНБ. гЮВЕЛ ? ЯЛ. ОПХЛЕП.

" The first step in creating an index on a large table now involves sorting the index key values. This significantly reduces the amount of i/o during the index build. Fast indexing is the default for tables with more than 500 rows and 30 OnLine data pages. To disable fast indexing and revert to the previous method for index creation you can set NOSORTINDEX as an environment variable before you begin your session. "

" Online 7.1 on a SPARC/1000; Solaris 2.4
<Trying to create an index on an integer field in a fragmented table
<(fragged on the int); If the table is approx 20K rows, engine wants about
<50 TBLSPACES during index creation;
<If the table is 200K rows, the engine wants about 400 TBLSPACES;
<We need 15 million rows;
<Engine initializes with TBLSPACES set at 500, but not with TBLSPACES set at
<600. I have not tried any larger values.
<The manual says nothing about (temporary) need for lots of TBLSPACES when
<creating an index on a fragged table.
<Any suggestions?
W're testing Online 7.1 on a Stratus FTX 2.3 platform.
The sizes of our tables range between 1.5 and 30 million rows.
We encountered the same problem in the case of a fragmented table
and a similar problem with an unfragmented table
(create index cannot allocate shared memory).
For both problems we use the following workaround. We set the undocumented
environment-variable NOSORTINDEX to true. (export NOSORTINDEX=true)"