This is the mail archive of the
mailing list for the systemtap project.
Re: IO script
- From: Mike Mason <mmlnx at us dot ibm dot com>
- To: Wenji Huang <wenji dot huang at oracle dot com>
- Cc: systemtap at sourceware dot org
- Date: Mon, 29 Oct 2007 11:22:20 -0700
- Subject: Re: IO script
- References: <4725A1C1.email@example.com>
Thanks for the scripts. Unfortunately, I don't think they measure actual physical seeks and random IO to disk. For that, I think we have to probe the IO schedulers. The scripts assume that physical IO's occur in the order requested, which isn't necessarily the case.
Also, as Srinivasa noted in bug 5231 today, bio_endio() no longer has a bytes_done variable as of 2.6.24-rc1 (Linus' git). We'll need to update the ioblock tapset and your scripts to reflect that.
Wenji Huang wrote:
Just as discussion at last conference call, Dtrace provides good
toolkit for users. see
If systemtap has same facility, it is very helpful to widen the use.
Based on idea from Dtrace, I created two scripts which can illustrate
disk I/O status.
Please kindly review it. If possible, we maybe collect/create some
useful stap scripts as toolkit embedded in source.