gstat batch mode

[spork]

Hi all,

I need to collect some stats on disk IO, and bsnmpd doesn't give me as much detail as I'd like (and I want to avoid net-snmp unless I really cannot). Mainly I want to look at tps, but %busy and other stuff gstat throws out would be interesting as well. I'm likely going to feed gstat output to a script and grab the output via snmp. Nagios will monitor this, but I'm really collecting the data for pnp so I can see some usage trends.

That said, when gstat is run in single-shot mode with the "-b" flag, does anyone know what data that represents? Is it a snapshot or an average over the last X seconds or what? The manpage isn't terribly clear on that.

If anyone wants to make an argument for iostat instead, I'm all ears. My main concern is that I'm getting somewhat accurate data.

Thanks!

 

[francis]

Hi, You wrote that bsnmpd does not giving you much details. But did you use it with the bsnmpd-ucd[1] module, which can monitor disk, but also CPU, network and memory? You should be able to get full information about disk with ucd module via snmp queries.

If anyone wants to make an argument for iostat...

There is also few commands to display I/O disk statistics. Some of them are]iostat 2[/CMD] # systat -iostat 1 Try to look here; iostat(8) (especially -x option), diskinfo(8).

[1] bsnmpd-ucd module is available in ports tree under net-mgmt/bsnmp-ucd category. So install it, then load usc module via /etc/snmpd.config file, and restart bsnmpd. Remember to put

bsnmpd_enable="YES"

in the /etc/rc.conf file. Maybe it will be useful for you?

 

[spork]

bsnmpd doesn't quite give me what I want, even with the ucd-snmp additions. I did end up settling on iostat instead, since I'm finding people claiming that if iostat is run twice, with say a 5 minute delay, the second set of stats returned will be an average of the last 5 minutes. This is perfect for our graphing. I just couldn't find a reference for gstat that described the batch stats.

Anyhow, this is what I'm pulling from iostat (in "extended" mode):

                        extended device statistics  
device     r/s   w/s    kr/s    kw/s wait svc_t  %b  
ada0      30.5   0.0  2974.7     0.0    1  11.0   5 
ada1      28.5   0.0  2367.0     0.0    1   2.6   3 
ada2      23.5   0.0  1855.2     0.0    0   2.8   2

My main interest is the tps on read and write and the number of outstanding transactions.

Hopefully this is helpful to someone else - lots of googling strongly suggests that setting a long delay and taking 2 samples will give the right information.