On Wed, Feb 4, 2009 at 2:22 AM, adrian cockcroft <adrian.cockcr...@gmail.com> wrote: > Orca is at http://www.orcaware.com > > collectors for most operating systems, saves into daily log files that are > easy to process with spreadsheets or use the orca system to turn them all > into a web site full of customizable rrdtool plots. I'm using it to collect > detailed data on some Linux systems right now... > > Other portable free tools that are more sophisticated: XEtoolkit, ganglia, > cacti > > Don't write yet another performance stats collector / plotter, its been done > to death. > > Adrian > > On Tue, Feb 3, 2009 at 7:57 PM, rickey c weisner <rohan...@comcast.net> > wrote: >> >> Jason, >> I have my own home grown script. It is named data_collector which >> is a ksh script controlled by an configuration file. It runs >> various stat tools plus a couple of tools of my own creation. >> >> It sticks all the data in a timestamped data directory. By default >> it runs for two hours, stops, and compresses the data. I usually >> put it in cron. >> >> Let me know if you want a copy and I'll find an ftp site for you or >> I can drop it into the email. It comes with the source for my tools. >> >> rick >> >> >> -- >> >> Rickey C. Weisner >> Software Development and Performance Specialist >> Principal Field Technologist >> Systems Quality Office >> cell phone: 615-308-1147 >> email: rick.weis...@sun.com >> _______________________________________________ >> perf-discuss mailing list >> perf-discuss@opensolaris.org > >
That's why I asked first :) I really don't want to reinvent the wheel if a good enough one already existed :) What would be easiest is if the data used by vmstat, mpstat, fsstat, etc was presented via snmp. That way there's one agent to run and that's it. I've been at places where there'd be 10+ different pieces of software that had to be installed for various forms of monitoring, all chewing up CPU (we'd only half joke we needed a dedicated CPU on each box just for the overhead of all of them). One in particular would run in the RT scheduling class which always made me uncomfortable. It looks like at one point there was a 'healthmonitor' module for the snmp agent that did a large amount of this, but appears it is no longer compiled or bundled. Would there be interest in reviving or creating a new module that could expose these statistics? AFAIK, the cmdline utilities all utilize kstats, so I imagine it wouldn't be too difficult. _______________________________________________ perf-discuss mailing list perf-discuss@opensolaris.org
