<tt>Mike Hamrick <mikeh@bluegecko.net> schrieb am 02.03.2008 14:13:51: > > You need to call iostat with multiple checks - the longer the better - > > but then it means you would have to run iostat for like 30 seconds or > > so -> plugin runtime is 30 seconds too then! That means that check > > would have a high delay/latency, which is overall a bad idea. My > > solution so far is to run the plugin via cron and report output via > > nsca, this gave me the best results. > > One thing you might want to consider is using sadc/sar for this job. > The sadc(8) program collects all kinds of stats on OS resource usage > and if you explicitly ask it, it will capture some interesting > disk i/o stats, including read/write requests per second. </tt> <tt>I actually have considered this, but found sadc not feasable for</tt> <tt>my intents. I am not interested in general iowait, but more in</tt> <tt>bandwidth usage of actual devices. Most of our high load servers</tt> <tt>have 3 SAN connections and it's important to know _which_ of those</tt> <tt>is under high stress and especially the performance data of</tt> <tt>those statistics to determine upcoming problems.</tt> <tt>At least I haven't been able to get sadc to log bandwitdh and</tt> <tt>throughput for each block device individually. sar in interactive</tt> <tt>mode is capable of doing it via "sar -dp 60 1" for example.</tt> <tt>Another problem that arises when using sadc is the problem that</tt> <tt>the scheduling of nagios is non predictable. I cannot garantuee,</tt> <tt>that nagios will poll the information once a minute for example.</tt> <tt>I cannot garantuee that nagios won't run the check twice in a</tt> <tt>longer checkperiod. The biggest feature of using it via send_ncsa</tt> <tt>passive checks, is the fact that I can inject the data when I want</tt> <tt>to - with perfect timing.</tt> <tt> But when using sar interactively, it's basically just a iostat -dp,</tt> <tt>which does the average calculation at the end - something my plugin</tt> <tt>already does ;)</tt> <tt>I'll put out our options in an abstract way:</tt> <tt>a) using sadc/sar:</tt> <tt>Have $script run by cronjob.</tt> <tt>Run active checks from nagios.</tt> <tt>b) using my passive plugin.</tt> <tt>Have $script run by cronjob.</tt> <tt>Let $script send output to nagios.</tt> <tt>In my opinion b) sound more appealing to me, since we have full</tt> <tt>control of timings. I do not see any benefit of using a).</tt> <tt>But maybe I am missing something - I am open to suggestions and</tt> <tt>ideas ;)</tt> <tt>S</tt> <tt>PS: Please do not CC me, I don't like dupes ;)</tt> -- Sascha Runschke Netzwerk- und Systemmanagement - Abt. Organisation und Betrieb - IT-Infrastruktur Telefon : +49 (201) 102-1879 Mobil : +49 (173) 5419665 Fax : +49 (201) 102-1102105 GFKL Financial Services AG Vorstand: Dr. Peter Jänsch (Vors.), Jürgen Baltes, Dr. Till Ergenzinger, Dr. Tom Haverkamp Vorsitzender des Aufsichtsrats: Dr. Georg F. Thoma Sitz: Limbecker Platz 1, 45127 Essen, Amtsgericht Essen, HRB 13522