[PVFS2-users] pvfs2-cp ignores the -s option and pvfs2-fs-dump hints how data aredistributed?

[Xiuduan Fang]

Hi all,

Any comments for my previous messages? In particular, I want to make sure if pvfs2-cp ignores the -s option. If this is the case, is there a patch for it? Thank you for your help.
  ----- Original Message ----- 
  From: Xiuduan Fang 
  To: pvfs2-users at beowulf-underground.org 
  Sent: Thursday, March 31, 2005 2:10 PM
  Subject: [PVFS2-users] pvfs2-cp ignores the -s option and pvfs2-fs-dump hints how data aredistributed?


  I installed pvfs2-1.0.1 on a cluster.  I tried to use the -s option of pvfs2-cp to change the stripe size but it appears that pvfs2-cp ignores the stripe size specified by -s.  I did some experiments as follows. First, I wrote a small C program to generate a file of a given size with the content of  "1a....a (1023 a's plus one 1 = 1K)2a...a (1023 a's plus one 2 = 1K)3a...a4a...". The content of the file gave the position information, i.e., the digital number k denoted the kth block (the block size is KBytes).  I used this program to create a file "/tmp/test_128K" of size 128K.  Then I ran "strace pvfs2-cp -s 131072 -t /tmp/test_128K /pvfs2/test128K 2>pvfs2cp" and the following shows some output of the strace debug file ,

    execve("/usr/centurion/bin/pvfs2-cp", ["pvfs2-cp", "-s", "131072", "-t", "/tmp/test_128K", "/pvfs2/test_128K2"], [/* 31 vars */]) = 0
  ¡­
  writev(4, [{"\277\312\0\0\10\0\0\0\r\0\0\0\0\0\0\0\0\0\1\0\0\0\0\0", 24}, {"1aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"..., 65536}], 2) = 65532
  writev(4, [{"aaaaaaaaaaaaaaaaaaaaaaaaaaaa", 28}], 1) = 28
  writev(6, [{"\277\312\0\0\10\0\0\0\16\0\0\0\0\0\0\0\0\0\1\0\0\0\0\0"..., 24}, {"65aaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"..., 65536}], 2) = 13032
  ...

  Based on the above writev commands, I guessed pvfs2-cp in the latest stable version 1.0.1always uses default stripe size 64KBytes. 

  Also I did some similar experiments to examine how data are distributed on PVFS2 I/O servers. I still used strace to trace the system calls called by pvfs2-cp and figured out that the file was simply round robin striped in some order like I/O server2, 3, 4, 5, and 1. Then I ran pvfs2-fs-dump and found pvfs2-fs-dump gave the same I/O server order for the file. It seems that pvfs2-fs-dump hints how data are distributed on I/O servers. If this is the case, then I can use striped size and the I/O server order provided by the output of pvfs2-dump to get the data distribution information.

  Please correct me if I am wrong. Thank you for any input.

  Xiuduan Fang,
  2nd Year Graduate, Computer Science Dept,
  University of Virginia,
  151 Engineer's Way, P.O. Box 400740,
  Charlottesville, VA 22904-4743
  Phone: 1-434-982-2296
  Email: xf4c at cs.virginia.edu



------------------------------------------------------------------------------


  _______________________________________________
  PVFS2-users mailing list
  PVFS2-users at beowulf-underground.org
  http://www.beowulf-underground.org/mailman/listinfo/pvfs2-users
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.beowulf-underground.org/pipermail/pvfs2-users/attachments/20050404/5e33e81c/attachment.htm