Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> VXFS an create datafile

VXFS an create datafile

From: <ylecornec_at_hotmail.com>
Date: 13 May 2005 02:58:05 -0700
Message-ID: <1115978285.566591.283740@g47g2000cwa.googlegroups.com>


Hi,
when i create a datafile on a VxFS FS and use vxtrace to see what happens while running the create statment: 100862 35477413 START write vdev voltest dg M01 dev 0,100020 block 6329184 len 512 concurrency 1 pid 1259
100863 35477413 START write vdev voltest dg M01 dev 0,100020 block 6329696 len 512 concurrency 2 pid 1259
100864 35477413 START write vdev voltest dg M01 dev 0,100020 block 6330208 len 512 concurrency 3 pid 1259
100865 35477414 START write vdev voltest dg M01 dev 0,100020 block 6330720 len 512 concurrency 4 pid 1259
100862 35477420 END write vdev voltest dg M01 op 100862 block 6329184 len 512 time 7
100863 35477420 END write vdev voltest dg M01 op 100863 block 6329696 len 512 time 7
100864 35477422 END write vdev voltest dg M01 op 100864 block 6330208 len 512 time 9
100865 35477424 END write vdev voltest dg M01 op 100865 block 6330720 len 512 time 10
100866 35477424 START write vdev voltest dg M01 dev 0,100020 block 6329184 len 2048 concurrency 1 pid 1259
100866 35477435 END write vdev voltest dg M01 op 100866 block 6329184 len 2048 time 11

what we can see here: VxFS makes two times the same write of 1Mo beginning at the block number 6329184 :
first write : 4 x 256Ko IOs (256Ko is value of write_pref_io) second write : 1 x 1Mo IO

discovered_direct_io seems to have a very strange behavior...has someone already noticed this behaviour, and understand what happens? this make create datafile statments last very long time.

Thanks Received on Fri May 13 2005 - 04:58:05 CDT

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US