| Disable Functions: Path : /usr/share/doc/dstat-0.7.2/ |
| Current File : //usr/share/doc/dstat-0.7.2/dstat-paper.txt |
= Dstat: pluggable real-time monitoring
Dag Wieers <dag@wieers.com>
$Id$
'This Dstat paper was originally written for LinuxConf Europe that was
held together with the Linux Kernel summit at the University in Cambridge,
UK in August 2007.'
== Introduction
Many tools exist to monitor hardware resources and software behaviour, but few
tools exist that allow you to easily monitor any conceivable counter.
Dstat was designed with the idea that it should be simple to plug in a piece
of code that extracts one or more counters, and make it visible in a way that
visually pleases the eye and helps you extract information in real-time.
By being able to select those counters that you want (and likely those
counters that matter to you in the job you're doing) you make it easier to
correlate raw numbers and see a pattern that may otherwise not be visible.
== A case for Dstat
A few years ago I was involved in a project that was testing a storage cluster
with a SAN back-end using GPFS and Samba for a broadcasting company. The
performance tests that were scheduled together with the customer took a few
weeks to measure the different behaviour under different stresses.
During these tests there was a need to see how each of the components behaved
and to find problematic behaviour during testing. Also, because it involved 5
GPFS nodes, we needed to make sure that the load was spread evenly during the
test. If everything went well repeatedly, the results were validated and the
next batch of tests could be prepared and run.
We started off using different tools at first, but the more counters we were
trying to capture the harder it was to post-process the information we had
collected. What's more, we often saw only after performing the tests that the
data was not representative because the numbers didn't add up. Sometimes it
was caused by the massive setup of clients that were autonomously stressing the
cluster. On other occasions we noticed that the network was the culprit. All in
all, we lost time because we could only validate the results by relating
numbers after the tests were complete and not during the tests.
Complicating the matter was the fact that 5 different nodes were involved
and using the normal command line tools like vmstat, iostat or ifstat (which
only showed us a small part of what was happening) was problematic as each
needed a different terminal. Besides, not all information was interesting.
Eventually Dstat was born, to make a dull task more enjoyable.
After the project was finished I was able to correlate system resources with
network throughput, TCP information, Samba sessions, GPFS throughput,
accumulated block device throughput, HBA throughput, all within a single
interval on one screen for the complete cluster.
== Dstat characteristics
There are many ideas incorporated into Dstat by design, and this section
serves to list all of them. Not all of them may appeal to the task you're
doing, but the combination may make it an appealing proposition nevertheless.
=== History of counters
An important characteristic in line-based tools like vmstat, iostat or
ifstat is the fact that you can compare historical collected data with
new data. This allows you to have a good feeling of how something is
evolving.
Compare this to tools like top or nmon, where data is often being refreshed
and you loose historical information (but in return can provide you with
a lot more information at the same time).
=== Adding unit indication
It was very important that when numbers were compared, they were in the same
unit, and not eg. a different power exponent. The human mind sometimes works
in mysterious ways and more so when working with numbers for hours and hours.
Adding the unit is something very convenient and may reduce the human error
factor.
Additionally, indicating the unit also makes sure that the columns have a
fixed width. Often when using vmstat or other tools, the columns tend to shift
depending on the width of the counter. This makes it very inconvenient to find
counters in the shifted output.
=== Colour highlighting units
After I added colours to help improve indicating units, I noticed that the
colours also helped to show patterns. This of course is very limited,
nevertheless it instantly shows when numbers are flat or changes are taking
place.
IMPORTANT: The colours are arbitrarily chosen. Do not make the mistake to
assume that green means good and red means bad. There is no real meaning to
the colour itself, however a change of colour does mean that a value has gone
over some pre-defined limit.
=== Intermediate updates
During tests, when you choose to see average values over a given time, it can
be useful to see how the averages evolve. Dstat, by default, displays
intermediate updates. This means that if you select to see 10 second averages,
after each second you see the accumulated average over the timespan. *This
means that after 4 seconds with intermediate updates, you see an average
taken over the 4 second timeframe.*
NOTE: This means that the closer you get to the given timeframe (eg. 10 seconds)
the more likely that it nears its final average over that period.
=== Adding custom counters
Dstat was specifically designed to enable anyone to add their own counters in a
matter of minutes. The plugin-based system takes care of displaying, colouring
and adding units to the counters. As a plugin-writer, you only have to focus
on extracting the counters from the kernel (procfs or sysfs), logfiles or
daemons.
=== Selecting plugins and counters
Being able to add custom counters is important, but selecting those counters
that you really need is even more important if you want to correlate counters
and see patterns. Less is more.
NOTE: In fact, Dstat currently does not allow you to select just counters, it
only allows you to select plugins. However, since you can modify or fork a
plugin, you still have the ability to select just those counters you prefer.
=== Exporting to CSV
Having information on screen is one thing, you most likely need some hard
evidence later to make your case. (Why else do all the work?)
Dstat allows to write out all counters in the greatest detail possible to CSV.
By default it also adds the command-line used for generating the output, as
well as a date and time stamp. Since Dstat in the first place is meant for
human-readable real-time statistics, it will by default also display the
counters to screen (unless you _/dev/null_ it).
TIP: Dstat appends to the output file so that you can add tests-results of
different tests to a single file. However, make sure that you tag each test
properly (eg. by using distinct filenames for each different test).
=== Time-plugin included
It may seem a small thing, but having exact time (and date) information for
your counters allows for a completely different usage as well. By adding
simple date and time information, Dstat can be used as a background process in
a screen to monitor the behaviour of your system during the night.
This proves to be very valuable for example, to find offending processes during
nightly tasks or to pinpoint their behaviour to certain events that you cannot
monitor during working hours.
It is also important when you have multiple Dstats running (eg. for nodes in a
cluster) to correlate counters between the outputs.
=== Terminal capabilities
Dstat also takes into account the width and height of your terminal window and
modifies output to fit into your terminal. This, of course, has no effect on
what ends up in the CSV output.
Another (debatable) useful feature is that Dstat will modify the terminal
title to indicate on what system it was run and what options were used.
Especially when monitoring nodes in a cluster, this can be useful, but even in
Gnome finding your Dstat window is handy.
WARNING: Some people however are annoyed by the fact that their distribution
does not reset the terminal title and Dstat therefor messes it up. There is no
way for Dstat to fix this.
== Plugins and counters
When we talk about plugins, we make a distinction between those plugins that
are included within the Dstat tool itself, and those that ship with it
externally. In essence there is no real difference, as the internal plugins
could easily have been created as an external plugin. The basic difference is
that the internal plugins have no dependencies except on procfs.
Having the basic plugins as part of Dstat, makes sure that Dstat can be moved
as a self-contained file to other systems.
=== Internal plugins
The plugins that have been selected to be part of the Dstat tool itself, and
therefor have no dependencies other than procfs, are:
- aio: asynchronous I/O counters
- cpu, cpu24: CPU counters (+-c+ and +-C+)
- disk, disk24, disk24old: disk counters (+-d+ and +-D+)
- epoch: seconds since Epoch (+-T+)
- fs: file system counters
- int, int24: interrupts per IRQ (+-i+ and +-I+)
- io: I/O requests completed (+-r+)
- ipc: IPC counters
- load: load counters (+-l+)
- lock: locking counters
- mem: memory usage (+-m+)
- net: network usage (+-n+ and +-N+)
- page, page24: paging counters (+-g+)
- proc: process counters (+-p+)
- raw: raw socket counters
- swap, swapold: swap usage (+-s+ and +-S+)
- socket: socket counters
- sys: system (kernel) countersA (+-y+)
- tcp: TCP socket counters
- time: date and time (+-t+)
- udp: UDP socket counters
- unix: unix socket counters
- vm: virtual memory counters
For backward compatibility with older kernels there is a cascading system that
selects the most appropriate internal plugin for your kernel. (eg. the
+dstat_disk+ plugin falls back to +dstat_disk24+ and +dstat_disk24old+) At this
moment there is no such system for external plugins.
=== External plugins
This basic functionality is easily extended by writing your own plugins
(subclasses of the python Dstat class) which are then inserted at runtime
into Dstat. A set of 'external' modules exist for:
- battery: battery usage
- battery-remain: remaining battery time
- cpufreq: CPU frequency
- dbus: DBUS connections
- disk-tps: disk transactions counters
- disk-util: disk utilization percentage
- dstat: dstat cputime consumption and latency
- dstat-cpu: dstat advanced cpu usage
- dstat-ctxt: dstat context switches
- dstat-mem: dstat advanced memory usage
- fan: Fan speed
- freespace: free space on filesystems
- gpfs: GPFS IO counters
- gpfs-ops: GPFS operations counters
- helloworld: Hello world dispenser
- innodb-buffer: innodb buffer counters
- innodb-io: innodb I/O counters
- innodb-ops: innodb operations counters
- lustre: lustre throughput counters
- memcache-hits: Memcache hit counters
- mysql5-cmds: MySQL communication counters
- mysql5-conn: MySQL connection counters
- mysql5-io: MySQL I/O counters
- mysql5-keys: MySQL keys counters
- mysql-io: MySQL I/O counters
- mysql-ops: MySQL operations counters
- net-mackets: number of packets received and transmitted
- nfs3: NFS3 client counters
- nfs3-ops: NFS3 client operations counters
- nfsd3: NFS3 server counters
- nfsd3-ops: NFS3 server operations counters
- ntp: NTP time counters
- postfix: postfix queue counters
- power: Power usage counters
- proc-count: total number of processes
- qmail: qmail queue sizes
- rpc: RPC client counters
- rpcd: RPC server counters
- sendmail: sendmail queue counters
- snooze: Dstat time delay counters
- squid: squid usage statistics
- thermal: Thermal counters
- top-bio: most expensive block I/O process
- top-bio-adv: most expensive block I/O process (advanced)
- top-cpu: most expensive cpu process
- top-cpu-adv: most expensive CPU process (advanced)
- top-cputime: process using the most CPU time
- top-cputime-avg: process having the highest average CPU time
- top-int: most frequent interrupt
- top-io: most expensive I/O process
- top-io-adv: most expensive I/O process (advanced)
- top-latency: process with the highest total latency
- top-latency-avg: process with the highest average latency
- top-mem: most expensive memory process
- top-oom: process first shot by OOM killer
- utmp: utmp counters
- vm-memctl: VMware guest memory counters
- vmk-hba: VMware kernel HBA counters
- vmk-int: VMware kernel interrupt counters
- vmk-nic: VMware kernel NIC counters
- vz-cpu: OpenVZ CPU counters
- vz-io: I/O usage per OpenVZ guest
- vz-ubc: OpenVZ user beancounters
- wifi: WIFI quality information
=== Most-wanted plugins
Hoping someone interested reads this document, I added a few plugins that
would be ``very nice'' to have but are currently lacking:
- slab: needs a VM expert to make sense out of the vast amount of data
- xorg: need information on how to get X resources, would be nice
to see evolution of X resources over time
- samba: lacking information to get counters from Samba without
forking smbstatus every second
- snmp: could be useful to relate counters from different systems
in a single Dstat
- topx: display the most expensive X application(s)
- systemtap: connecting Dstat to systemtap counters
Creative souls with other ideas are welcome as well !
== Using Dstat
Central to the Dstat command line interface is the selection of plugins. The
selection and order of options influence the Dstat output directly.
=== Enabling plugins
The internal plugins have short and/or long options within Dstat, eg. +-c+ or
+--cpu+ will enable the cpu counters.
The external plugins are enable by a long option including their name,
eg. +--top-cpu+
The following examples will enable the time, cpu and disk plugins, and are
equal.
----
dstat -tcd
dstat --time --cpu --disk
----
=== Total or individual counters
Some of the plugins can show both total values or individual values and
therefor have an extra option to influence this decision.
----
dstat -d -D sda,sdb
dstat -n -N eth0,eth1
dstat -c -C total,0,1
----
You can show both the individual values and total values as follows:
----
[dag@horsea ~]$ dstat -d -D total,hda,hdc
-dsk/total----dsk/hda-----dsk/hdc--
read writ: read writ: read writ
1384k 1502k: 114k 1332k: 81k 359B
0 44k: 0 44k: 0 0
0 0 : 0 0 : 0 0
----
The special +-f+ or +--full+ option allows to select individual counters by
default, and can be overruled by +-C+, +-D+, +-I+, +-N+ or +-S+.
=== Influencing output
Dstat has a few more options to influence its output. With the +--nocolor+
one can disable colours. The +--noheaders+ option disables repeating headers.
The +--noupdate+ option disables intermediate updates. The +--output+ option
is used for writing out to a CSV file.
=== Plugin search path
Dstat looks in the following places for plugins. This allows a user without
root privileges to use some extra plugins.
- ~/.dstat/
- <binarypath>/plugins/
- /usr/share/dstat/
- /usr/local/share/dstat/
The option +--list+ shows the available plugins and their location in the
order that the plugin search path is used.
NOTE: Plugins are named +dstat_<name>.py+.
== Use-cases
Below are some use-cases to demonstrate the usage of Dstat.
WARNING: The following examples do not look as nice as they do on screen
because this document is not printed in colour (and I did not prepare it in
colour :-)).
=== Simple system check
Let's say you quickly want to see if the system is doing alright. In the past
this probably was a +vmstat 1+, as of now you would do:
----
dstat -taf
----
.Sample output
----
[dag@rhun dag]$ dstat -taf
-----time----- -------cpu0-usage------ --dsk/sda-----dsk/sr0-- --net/eth1- ---paging-- ---system--
date/time |usr sys idl wai hiq siq| read writ: read writ| recv send| in out | int csw
02-08 02:42:48| 10 2 85 2 0 0| 22k 23k: 1.8B 0 | 0 0 |2588B 2952B| 558 580
02-08 02:42:49| 4 3 93 0 0 0| 0 0 : 0 0 | 0 0 | 0 0 |1116 962
02-08 02:42:50| 5 2 90 0 2 1| 0 28k: 0 0 | 0 0 | 0 0 |1380 1136
02-08 02:42:51| 11 6 82 0 1 0| 0 0 : 0 0 | 0 0 | 0 0 |1277 1340
02-08 02:42:52| 3 3 93 0 1 0| 0 84k: 0 0 | 0 0 | 0 0 |1311 1034
----
NOTE: The +-t+ here is completely optional and generally wastes space. But
often you are not monitoring for 10 seconds but rather measure in minutes or
hours. Having a general idea on what timescale counters have been averaged is
nevertheless interesting.
=== What is this system doing now ?
I often run both the +dstat_top_cpu+ and +dstat_top_mem+ programs on a system,
just to see what a system is doing. Having a quick look at what application
is using the most CPU over a few minutes and to see what the general usage
of memory is of the top application gives away a lot about a system.
.Sample output
----
[dag@horsea dag]$ dstat -c --top-cpu -dng --top-mem
----total-cpu-usage---- -most-expensive- -dsk/total- -net/total- ---paging-- -most-expensive-
usr sys idl wai hiq siq| cpu process | read writ| recv send| in out | memory process
9 2 80 9 0 0|kswapd 0| 123k 164k| 0 0 |9196B 18k|rsync 74M
2 3 95 0 0 0|sendmail 1| 0 168k|2584B 39k| 0 0 |rsync 74M
18 3 79 0 0 0|httpd 17| 0 88k|5759B 118k| 0 0 |rsync 74M
3 2 94 1 0 0|sendmail 1|4096B 0 |2291B 4190B| 0 0 |rsync 74M
2 3 95 0 0 0|httpd 1| 0 0 |2871B 3201B| 0 0 |rsync 74M
10 7 83 0 0 0|httpd 13| 0 0 |2216B 10k| 0 0 |rsync 74M
2 2 96 0 0 0| | 0 52k| 724B 2674B| 0 0 |rsync 74M
----
=== What process is using all my CPU, memory or I/O at 4:20 AM ?
Imagine the monitoring team notices strange peaks, a system engineer got a
worthless message, the system was swapping extensively, a process got killed.
Something indicates the system is doing something unexpected but what is
causing it and why ? As of now you can do:
----
screen dstat -tcy --top-cpu 120
screen dstat -tmgs --top-mem 120
screen dstat -tdi --top-io 120
----
to see what process is using the most CPU, the most memory and the most I/O
resources.
And hopefully one day we can do:
----
dstat -tn --top-net 120
dstat -tn --top-x 120
----
Leave it running during the night and in the morning you can see the light.
=== How much ticks per second on my kernel ?
In some cases it can be useful to see how many ticks (timer interrupts) your
kernel is producing. With older kernels this is a fixed number (usually 100,
250 or 1000) but on newer kernels the number can be dynamic.
Also on VMware virtual machines, the number of ticks can cause clock issues,
so in that case if you want to see what is happening, you can simply do:
----
dstat -ti -I0 --snooze --debug
----
Dstat nowadays can also detect lost ticks (when the number of ticks do not
match the time progress. This is useful to correlate VM issues with other
problems.
////
=== Monitoring memory consumption of a process over time
Now, I have twice used Dstat to verify memory usage. And I have concluded that
2 programs have severe memory leaks. One, unsurprisingly, is Firefox, the
other sadly is wnck-applet (yes, unfortunately).
Now Dstat is currently not really useful for specifying your own process to
monitor (unless you dig into the module, which is easier than one might
expect). But I am already anticipating Pstat, which is a Dstat but for
process-related counters.
More on this later...
////
=== What device is slowing down my system ?
A nice feature of Dstat is that it can show how many interrupts each of your
devices is generating. The 'cpu' stats already show this in percentage as
'hard interrupt' and 'soft interrupt', and the 'sys' stats shows the total
number of interrupts, but the 'int' stats go into detail. And you can specify
exactly what IRQs you want to watch.
Many devices generate interrupts, especially when used at maximum capacity.
Sometimes too many interrupts can slow down a system. If you want to correlate
bad performance with hardware interrupts, you can run a command like:
----
dstat -tyif
dstat -tyi -I 12,58,iwlagn -f 5
----
Much like +watch -n1 -d cat /proc/interrupts+ on steroids.
----
dstat -t -y -i -f
----
which then results in:
.Sample output
----
[dag@rhun ~]$ dstat -t -y -i -f 5
-----time----- ---system-- -------------------interrupts------------------
date/time | int csw | 1 9 12 14 15 58 177 185
13-08 21:52:53| 740 923 | 1 0 18 5 1 17 4 131
13-08 21:52:58|1491 2085 | 0 4 351 1 2 37 0 97
13-08 21:53:03|1464 1981 | 0 0 332 1 3 31 0 96
13-08 21:53:08|1343 1977 | 0 0 215 1 2 32 0 93
13-08 21:53:13|1145 1918 | 0 0 12 0 3 33 0 95
----
When having the following hardware:
----
[dag@rhun ~]$ cat /proc/interrupts
CPU0
0: 143766685 IO-APIC-edge timer
1: 374043 IO-APIC-edge i8042
9: 102564 IO-APIC-level acpi
12: 4481057 IO-APIC-edge i8042
14: 1192508 IO-APIC-edge libata
15: 358891 IO-APIC-edge libata
58: 4391819 IO-APIC-level ipw2200
177: 993740 IO-APIC-level Intel ICH6
185: 33542364 IO-APIC-level yenta, uhci_hcd:usb1, eth0, i915@pci:0000:00:02.0
NMI: 0
LOC: 143766578
ERR: 0
MIS: 0
----
Or select specific interrupts:
----
dstat -t -y -i -I 12,58,185 -f 5
----
Another possibility is to use the +--top-int+ plugin, showing you the most
frequent plugin on your system:
----
[dag@rhun ~]# dstat -t --top-int
----system---- ---most-frequent----
time | interrupt
11-06 08:34:53|ahci 5
11-06 08:34:54|i8042 69
11-06 08:34:55|i8042 45
11-06 08:34:56|ehci/usb2 12
11-06 08:34:57|
----
=== How does my WIFI signal evolve when I move my laptop or AP through the house ?
Something I was looking into when trying to find the optimal location for the
WIFI access point. However I must say that another tool I wrote 'Dwscan' is
currently more sophisticated.
----
dstat -t --wifi
----
=== Is my SWRAID performing as it claims ?
You can monitor I/O throughput for any block device. By default dstat limits
itself to real block devices to prevent having the same I/O to be counted more
than once, but if you want to monitor a SWRAID device, or a multipath device,
you can simply do that by doing:
----
dstat -td -D md0,md1,sda,sdb,hda
----
== Writing your own Dstat plugin
Dstat is completely written in python and this makes it extremely convenient
to write your own plugins. The many plugins that come with Dstat are an
excellent source of information if you want to write your own.
=== Introducing the hello world plugin
The following plugin does nothing more than write "Hello world!" to its
output.
.The dstat_helloworld plugin in its full glory.
----
class dstat_helloworld(dstat):
``"
Example "Hello world!" output plugin for aspiring Dstat developers.
``"
def __init__(self):
self.name = 'plugin title' <1>
self.nick = ('counter',) <2>
self.vars = ('text',) <3>
self.type = 's' <4>
self.width = 12 <5>
self.scale = 0 <6>
def extract(self):
self.val['text'] = 'Hello world!' <7>
----
In this example, there are several components:
1. +self.name+ contains the plugin's visible title.
2. +self.nick+ is a list of the counter names
3. +self.vars+ is a list of the variable names for each counter
4. +self.type+ defines the counter type: string, percentage, integer, float
5. +self.width+ defines the column width
6. +self.scale+ influences the coloring and unit type
7. +self.val+ contains the counter values that are being displayed
=== Parsing counters
The following example shows how information is collected and counters are
processed. It also includes a +check()+ method to properly bail out when the
system fails to meet some plugin criteria.
.The dstat_postfix plugin
----
class dstat_postfix(dstat):
def __init__(self):
self.name = 'postfix'
self.nick = ('inco', 'actv', 'dfrd', 'bnce', 'defr')
self.vars = ('incoming', 'active', 'deferred', 'bounce', 'defer')
self.type = 'd' <1>
self.width = 4
self.scale = 100
def check(self): <2>
if not os.access('/var/spool/postfix/active', os.R_OK):
raise Exception, 'Cannot access postfix queues'
def extract(self):
for item in self.vars: <3>
self.val[item] = len(glob.glob('/var/spool/postfix/'+item+'/*/*')
----
This example shows the following items:
1. type, width and scale specify decimal, column width a,d coloring based on
multiplication of 100
2. The +check()+ method tests conditions and bails out of they are not met
3. To make processing easier we have opted to use as value names (+self.vars+)
the name of the postfix queues and store counts in +self.val+
=== Opening files
Dstat provides its own +dopen()+ function to plugins. Using +dopen()+ instead
of +open()+ plugins do not need to reopen files to update their counters. But
this is only useful when plugins open a few files. For eg. opening _/proc/pid_
files the number of open files would only be increasing as the number of
processes increases.
=== Piping to an application
Dstat provides its own +dpopen()+ function to plugins. This function allows
the plugin to open stdin, stdout and stderr pipes for 2-way communication with
processes. To see this in action, take a look at the +dstat_gpfs+ plugins or
the +dstat_mysql+ plugins.
Piping to an application is more expensive than getting kernel counters from
_/proc_, but it beats having to run a program and capturing the output.
== Known issues
There are some known issues that are important to understand when using Dstat.
=== Writing Dstat and plugins in C
It makes sense to reimplement Dstat or some of its plugins in C and still
allow the writing of Python (or even Perl) plugins. Tests have shown that for
example processing _/proc/pid_ in C makes the plugin 3 times faster. And this
did not take into account the processing of the results and displaying the
output.
So rewriting in C makes a lot of sense, but it is also much more complicated.
=== Python 1.5
There used to be a Python 1.5 version of Dstat, but with RHEL2 going out of
support in 2009 I decided to no longer spend the extra effort to sync and test
the Dstat15 version.
Leaving Python 1.5 behind means that plugins do not longer have to be
compatible with Python 1.5 either. It is no coincedence that after this event
a major overhaul was made to the plugin interface.
=== Counter rollovers
Unfortunately Dstat is susceptible for counters that ``rollover''. This means
that a counter gets bigger than its maximum value the data-structure is capable
of storing. As a result the counter is reset.
For some architectures and some counters, Linux implements 32bit values, this
means that such counter can go up to 2^32 (= 4294967296B = 4G) values.
For example the network counters are calculated in absolute bytes. Every 4GB
that is being transferred over the network will cause a counter reset. For
example on a bonded 2x10Gbps interfaces that is using its theoretical transfer
limit, this would happen every 1.6 seconds.
Since _/proc_ is updated every second, this would be impossible for Dstat to
catch. Currently if Dstat encounters a negative difference for an interval it
assumes a single rollover has happened and compensates for it. If that
assumption is wrong, the user is working with wrong counters nonetheless.
If you suspect that the behaviour of your system is susceptible of counter
rollovers, make sure you take this into account when using Dstat (or any other
tool that uses these counters for that matter).
TIP: Shipped with the Dstat documentation there is a document
(_counter-rollovers.txt_) that goes deeper into counter rollovers. If this
affects you, read that document and contact me for possible implementation
changes to improve handling them.
== Dstat performance
As mentioned several times now, Dstat is written in python. There are various
reasons that Python was chosen and the most important reason is that we target
system engineers and users, so we need to simplify writing plugins, processing
counters and lowers the bar for people to contribute changes.
The downside of choosing a scripting language is that it is slower than if it
would be written in C, obviously. *Dstat is not optimised for performance.*
NOTE: This may seem ironic: a performance monitoring tool that is not
optimised for performance, but rather for flexibility. However the ease of
writing plugins and prototyping gets precedence over performance at this time.
On the other hand we have pretty good tools to measure the overhead of a
single plugin and profiling infrastructure to counter any excuses for sloppy
plugin development.
=== Plugin performance
If we look at the basic plugins, there are no real performance issues with
Dstat. Loading Dstat takes longer to start than eg. vmstat, but once running,
Dstat's performance for the same functionality is up to par with vmstat,
ifstat and other similar tools.
However there are *some plugins that are much more resource intensive than
others* and the selection of plugins determines Dstat's performance in a major
way.
=== Performance monitoring Dstat
Dstat comes with some plugins (starting with +dstat_+) to check the overhead
of itself, this together with the selection of plugins makes it very
convenient to measure the overhead of individual plugins. The following
options exist (as plugins):
--dstat::
Provides cputime and latency information for Dstat. This plugin can help you
determine how accurate and how much overhead Dstat has with its current
plugins enabled.
--dstat-cpu::
Provides cpu utilization (user-space and kernel-space) statistics for Dstat.
This plugin can help determine where there is some room for improvement for
individual plugins (or Dstat itself).
--dstat-ctxt::
Provides context switch information for Dstat. Both voluntary as well ass
involuntary context switches are shown, providing you with some idea of how
the system is providing timeslices and how Dstat is returning the cpu to the
system.
--dstat-mem::
Provides memory information about the Dstat process. This plugin enables
plugin developers to determine whether Dstat is increasing its memory usage
and therefor is 'leaking' memory over time. This plugin proved very useful in
optimizing memory usage of the top-plugins, which typically scan all
processes.
--snooze::
This plugin shows in milliseconds how much time is deviating from the previous
run. Which is influenced by the time it takes for earlier stats to be
calculated. So the output of this plugin is very dependant on the location on
the command-line.
--debug::
This option is not a plugin, but internal to Dstat. It will cause Dstat to
show the actual time in milliseconds from start to end at the end of each
line. This should be more or less close to the output of the +dstat_dstat+ and
+dstat_dstat_cpu+ plugins.
+
It also influences the internal +dstat_time+ plugin to show milliseconds
instead of seconds, which may help showing the accuracy of Dstat itself.
--profile::
Ths option is also not a plugin, but internal to Dstat. It provides you with
detailed profiling information at the end of each run. The default settings
can be changed inside Dstat (or a copy) to tweak the output you are looking
for. It creates a termporary profiling file in the current directory when
running, but will clean it up after exit.
// FIXME: Please improve the examples by using the --dstat plugins
=== Measuring plugins
Here is a small example of how one can measure the impact of a plugin.
.The cost of running the timer plugin
----
[dag@rhun dag]$ dstat -t --debug
Module dstat_time
-----time-----
date/time
19-08 20:34:21 5.90ms
19-08 20:34:22 0.17ms
19-08 20:34:23 0.18ms
19-08 20:34:24 0.18ms
----
Compare this with other plugins to see what the cost is of an individual
plugin.
.The cost of running the +dstat_cpu+ plugin
----
[dag@rhun dstat]$ dstat -c --debug
Module dstat_cpu requires ['/proc/stat']
----total-cpu-usage----
usr sys idl wai hiq siq
15 3 77 4 0 1 11.07ms
5 3 92 0 0 0 0.66ms
5 4 91 0 0 0 0.65ms
5 3 92 0 0 0 0.66ms
----
As you can see, getting the CPU counters and calculating the CPU usage takes
up 0.5 milliseconds on this particular system. But if we look at the usage of
the +dstat_top_cpu+ plugin:
.The cost of running the +dstat_top_cpu+ plugin
----
[dag@rhun dstat]$ dstat --top-cpu --debug
Module dstat_top_cpu
-most-expensive-
cpu process
Xorg 2 43.82ms
Xorg 1 33.23ms
firefox-bin 2 33.54ms
Xorg 1 33.24ms
----
we see that processing the _/proc/pid_ files causes the top-cpu plugin to use
an additional 33ms.
WARNING: These values show the time it takes to process the plugins and does
not indicate the amount of CPU usage Dstat consumes. This obviously means that
the process time of plugins depends on how much the system is being stressed
as well as on what the plugin exactly is doing.
Plugins that communicate with other processes or those that process lots of
information (eg. communicating with the mysql client, or processing the mail
queue) may not actually use any local resources, but the latency causes
Dstat to slow down processing other counters.
// FIXME: Write about profiling infrastructure
== Future development
The Dstat release contains a _TODO_ file highlighting all the items and
ideas that have been played with. Here is a list of the most important ones:
- Output
* Changes in how Dstat colours digits within a value (the 6 in 6134B)
- Exporting information
* Connecting Dstat with rrdtool
* Exporting to syslog or remote syslog (a way to transport counters ?)
- Plugins
* Be smart when plugins are loaded more than once (some plugins could
benefit)
* Add more plugins
- Redesign Dstat
* Create an object-model and namespace for plugins and counters so that
other tools can be based on Dstat
== Links
- http://dag.wieers.com/home-made/dstat/[Dstat homepage]
- http://svn.rpmforge.net/svn/trunk/tools/dstat/[Dstat subversion]
- http://lists.rpmforge.net/mailman/listinfo/tools[Dstat mailinglist]
// vim: set syntax=asciidoc: