EDM01-02: DAG 4.2S Card User Guide
©2005
21
Version 7: May 2006
Process
, continued
Process
Description
Starting a capture
session.
Once the capture parameters are configured, a
capture session is started by:
tools/dagsnap –d dag0 –v –o tracefile0 &
tools/dagsnap –d dag1 –v –o tracefile1
Option
-v
provides user information during
capture; it can be omitted for automated trace
runs.
If the
tracefile
parameter is not specified the
tool writes to
stdout
, which can be used to
pipeline
dagsnap
with other tools from
dagtools
package.
By default
dagsnap
runs forever.
Stopping
dagsnap
can be stopped with a signal:
killall dagsnap
,
or key strokes CTL + C
dagsnap
can also be configured to run for a fixed
number of seconds and then exit using the
–s
option.
5.2 High Load Performance
Description
As the DAG card captures packets from the network link, it writes a
record for each packet into a large buffer in the host PC’s main memory.
Avoiding
packet loss
In order to avoid packet loss, the user application reading the record, such
as
dagsnap
, must be able to read records out of the buffer faster than they
arrive. Otherwise the buffer eventually fills and packet records are lost.
For Linux and FreeBSD, when the PC buffer becomes full, the message:
kernel: dagN: pbm safety net reached 0xNNNNNNNN”
is displayed on the PC screen, and printed to
log /var/log/messages
.
The “Data capture” LED also goes out. This may be visibly indicated as
flashing or flickering.
