 |
 |
Archives of the TeradataForum
Message Posted: Tue, 19 Jul 2005 @ 13:13:23 GMT
| Subj: | | Re: FastExport to a named pipe |
| |
| From: | | Geoffrey Rommel |
| | When it fails, it always fails at the same point (same total number of bytes read). | |
This suggests that the user running the FastExport has a ulimit for pipes but not for files. The command "ulimit -a" will tell you the limits.
Note that when ulimit says "blocks", it means sectors of 512 bytes = 0.5 KB.
| | We have experimented with various block sizes and can find no commonality, for example it will fail with the default BLOCKSIZE
(64260) or with BLOCKSIZE of 62000, but succeed with 32000 or 64000, or even 64010. | |
This is a bit of a puzzler, but here's an idea. Run 'stat' on the named pipe and find the preferred blocksize. Your C programmer can do this,
or run this Perl script:
#!/usr/bin/perl
$FI = shift;
@st = stat($FI);
print "Preferred blocksize is $st[11]\n";
On both MP-RAS and Solaris, I get 5120 for a named pipe. Then just use a multiple of this size -- 61440, for instance.
| |
