File corruption during IOZONE tests with 8GB or larger test files
I am having a problem running the IOZONE disk test program on a new storage setup. I think I maybe getting corrupted files from the storage system but I need to confirm it is not a problem with IOZONE itself. There is not much info on IOZONE other than the main web page and man pages.
I need to run rather large test files as the server host in used for the test has 16 GB of RAM.
An example run was to start the program in the working directory of the filesystem I wanted to test:
/opt/iozone/bin/iozone -Ra -g 20G -b /root/iz_20G.txt
Program starts up as normal:
Iozone: Performance Test of File I/O
Version $Revision: 3.291 $
Compiled for 64 bit mode.
(extra details removed)
Run began: Sat Apr 18 07:12:16 2009
Excel chart generation enabled
Using maximum file size of 20971520 kilobytes.
Command line used: /opt/iozone/bin/iozone -Ra -g 20G -b /root/iz_20G.txt
Output is in Kbytes/sec
Time Resolution = 0.000001 seconds.
Processor cache size set to 1024 Kbytes.
Processor cache line size set to 32 bytes.
File stride size set to 17 * record size.
The tests runs fine until it gets into the 8 GB file range.
and then it abruptly ends:
Error in file: Found ?0? Expecting ?2020202020202020? addr 2aaaaab20000
Error in file: Position 2531393536
Record # 9656 Record size 256 kb
where 2aaaaab20000 loop 131072
I have tried this three times and got the failure when the test file reached 8GB twice and once when I started with a 4 GB files that stopped at a 16 GB test file. I am only running one test at a time, so I do not have multiple iozone instances trying to use the same file.
I have also used the "-f" switch to name a file for use vs. filename generated by the iozone program itself and had the same problem.
I am about to send of details to the vendor of the storage solution.
However, I want to rule out that this is not a problem with the IOZONE program. I have ran similiar tests using IOZONE on local filesystems, and have not seen this problem. I have also tried two versions of IOZONE (iozone-3-291 and iozone-3-318 both 64bit).
If anyone can confirm that the above errors indicate a corrupted test file that would be most helpful.
If anyone has seen this before and can confirm it is a possible bug in iozone that would be great also.
I have not tried the latest version, I will do that soon.
Thanks in advance if anyone has any thoughts on this problem.