On Sun, Mar 1, 2009 at 4:19 PM, Peter Tribble <peter.tribble at gmail.com> wrote: > On Fri, Feb 27, 2009 at 4:37 AM, Jason King <jason at ansipunx.net> wrote: >> As one may have noticed, I've uploaded an initial MIB into hg for the >> purposes of capturing the discussion that's happened so far. ?It is in >> no way complete and should not be viewed as such. > > It may help to point out that you can use OpenGrok to browse to the files: > > http://src.opensolaris.org/source/xref/perfmib/ > > Err, Coutner64?
Fixed (not sure how often OpenGrok updates) -- until we can get an actual OID, I'm not sure of a good way to validate it (and thus be able to catch typos such as that except by inspection). > > Should the fsstat numbers allow you to have the filesystem types as well > as mountpoints? You mean aggregrate by filesystem type? That could be added, though the type is listed, so it could also be aggregrated externally. I've personally never found much use in aggregating by filesystem type (most systems I've dealt with tend to be UFS /, vxfs everything else) over having it broke down by individual filesystems, however if others have had found it useful, I'm happy to defer to their experiences. > > The disk stats should include the queues too. And what about devices that > don't Are we sure the queue data is the right thing to publish? I think there was acknowledgement that it can be hard to interpret (or rather easy to misinterpret -- at least in relation to other metrics commonly used). I would like to dig deeper (hopefully with others giving insight into their experiences) as to the usefulness of the queue data versus something else (even if it doesn't exist right now). It is easier to add to the MIB than to remove, so I want to be confident about what is added. > export the error kstats that supply half the data? (Some drivers > didn't until recently, > and things like metadevices don't either. Are there other things that > might export > statistics that look like IO stats?) I'm not sure what you mean -- you mean devices that aren't providing error kstats (or aren't providing errors to the sd, ssd, etc. drivers to have their kstats updated)? If so, I would think we should file bugs (or if the drivers are open, submit a patch as well).
