Testing AFP with Bonnie++

[amsterdam]

I’m looking to implement an iSCSI-based backend storage system for our next server rollout this summer. I want to gather some benchmarking data to compare how the new setup would perform compared to the current one.

My plan is to use Bonnie++ to test our current network fileshares, but I’m running into problems. I can compile Bonnie++ using Macports and run it on the local filesystem, but when I try to run a small test on a AFP share, Bonnie++ reports:

[code]
$ /opt/local/bin/bonnie++ -s 10M -d [AFP share]
Writing a byte at a time…done
Writing intelligently…done
Rewriting…done
Reading a byte at a time…done
Reading intelligently…done
start ’em…done…done…done…done…done…
Create files in sequential order…done.
Stat files in sequential order…done.
Delete files in sequential order…Bonnie: drastic I/O error (rmdir): Directory not empty
Cleaning up test directory after error.
[/code]

Has anyone been able to use Bonnie++ to test their network filesystems?

Here are some notes:
1. Bonnie, the original version from the 1990s works fine on the AFP shares, but doesn’t have as robust a feature set
2. I tried testing on an AFP share that was world-writeable, so no permissions issues
3. Can’t find any errors in the logs on either server or client
4. Current system is a 10.5.6 G5 Xserve with an XRAID

Any help is appreciated, as well as other ideas for benchmarking/testing.

thanks!

[amsterdam]

It seems the problem was because Bonnie++ likes to create 1024 zero-length files as part of its test. It puts these in one sub-directory. However, if I set the test to split these files out in 100 different sub-directories, it seems to work fine. Perhaps a limitation with AFP?

[Author]

Posts