Frequently asked questions
What is Collect?
Collect is a tool for collecting operating system and process
information for all supported versions of Tru64 UNIX.
Where can I get Collect?
Collect is now shipped on the Tru64 UNIX base operating system,
versions 5.0A and later, and 4.0G. Collect is also available
in the latest patch kits for Tru64 UNIX versions 5.0, 4.0D,
4.0E, and 4.0F.
What other sources of information for Collect
exist?
Collect, collgui and cfilt all
have very informative reference pages included in the setld installation.
Each has a brief usage message available using the -h option.
You may also try the Collect website for
more information, or join the Tru64 UNIX forum at www.tru64.org/phorum.
Please understand that this forum is not supported by HP.
Are there any tools to help visualize and
filter the mass of information that Collect gathers?
The following adjunct tools exist to help users better use
Collect data:
collgui - a graphical front-end for
Collect
cfilt - a filter for Collect
Who are Collect's customers?
Major users of Collect are MCI, Nasdaq, Department of Defense,
T-Mobil, Deutsche Bank, AOL, Lycos, and Yahoo.
Can Collect survive a reboot?
Yes. On Tru64 UNIX 5.0A and higher systems, Collect will
restart if you configure your system with this command:
# rcmgr COLLECT_AUTORUN 1
If collgui and cfilt are
written in Perl and Perl/Tk respectively, can I filter
and visualize Collect datafiles on a PC under Windows?
No, because cfilt implicitly uses Collect to
read datafiles before applying filters and passing the data
on to gnuplot or Excel. You can, however, pre-convert your
datafiles on a Tru64 UNIX system by using cfilt (implicitly
using Collect) to convert to gnuplot or Excel format.
You can also pre-convert your Collect compressed
files to stdout format (human readable),
which can be quite large because of all the empty space.
Then run cfilt on them using a Perl interpreter
on a Windows PC. You have to figure out the cfilt incantation
yourself instead of having collgui do
it for you, because collgui was designed
for X11 and doesn't present well on a PC.
Why do people use Collect?
It is very lightweight (it uses less than 1% of CPU capacity
at a 30-second sampling interval).
It gathers as much performance data as reasonably
possible (and it can collect data at sub-second sampling
rates if necessary).
It is highly flexible:
- It can be configured to collect all or any combination of subsystems.
- It can run in either interactive or historical mode.
- When coupled with
collgui and cfilt, Collect can display graphs from either real-time or pre-collected data.
Collect crashes with the following output:
readk: No such device or address
FATAL: readk: Can't read from kernel
Collect gathers all its data from the kernel and needs to
know certain kernel symbol names to access that data. As
the kernel evolves, those symbol names may change and cause
difficulty for Collect. To remedy this, make sure you have
the latest version of Collect; that is, the one that came
with the Base OS on which you are running.
Collect fails to read the output datafile
with the following error:
Data version newer than
collect version - you may have
problems
The output datafile was created by a newer Collect binary,
so you need to obtain a newer Collect binary than you currently
have. This is available from the OS patch kit or the Collect
FTP site.
How can I use Excel to plot or graph Collect
datafiles?
You can use either collgui or cfilt to
export to Excel:
Using collgui:
- Run
collgui in debug mode:
>> collgui -d "collect datafile"
- Select the desired subsystems and click on Display.
- Return to the shell that
collgui was
started in. You will see that collgui has
created a file in the /var/tmp directory.
The filename is collgui.xxxx, where xxxx are
integers. This datafile (collgui.xxxx)
is exportable to Excel. Copy it to a Windows system.
- On your Windows system, start Excel and open
collgui.xxxx. You
might have to change the Files of Type: field to "All
files(*.*)".
- In Excel 2000, a text import wizard will pop up.
- In Excel 2000, select data type Delimited, then
select Next.
- In Excel 2000, select Tabs and Spaces as Delimiters,
then select Next.
- In the Data Preview pane, select the columns that
you want to import using the Shift key, then select
Finish.
- You should now see the columns displayed in your
worksheet.
Using cfilt:
- Use
cfilt to generate the datafile.
For example, if you choose to display the system
time, physical memory used for the process
data, and user+system time, wait time for
the Single CPU field, enter the following command:
cfilt -f "collect datafile" 'sin:WAIT:USER+SYS'
'pro:Systim#:RSS#'
> /var/tmp/collgui.xxxx.
Copy the collgui datafile to your Windows
system.
- Follow steps 4-9 in the Using
collgui procedure
above.
What is the difference between Normalize
and Raw Data?
In raw mode, each point in the data set represents a raw
(unmodified) value.
In normalized mode, the point of biggest value in
the data set is normalized to 100. Other points in
this data set are then scaled down from 100, proportionally.
For example:
Given data set [4,20,6,8,200]
Raw mode [4,20,6,8,200]
Normalized mode [2,10,3,4,100]
As a result, when plotting in normalized mode, the
Y axis always has values in the range from 0 to 100.
Because a normalized graph does not show raw or absolute
values, its main use is to show a trend. To see exact
or raw values, plot in raw mode.
When viewing my graph, what is the scale
used for the Y axis, and how should it be interpreted?
You can use the context of the graph to induce the unit that's
being used. For example, if the graph is showing CPU usage
such as user time, idle time,... the Y axis
is the percentage. A more direct and accurate way to know
the exact value of each metric is to play back the datafile
using collect as in the following example:
>> collect -p datafile.cgz
More Collect product details
|
