To capture data on a remote system, run blktrace
with different options on both the system that stores the data and
the system to be traced.
Procedure
- On the system where the captured data should be stored
start blktrace in server mode. This system should have a good network
connectivity to the system being traced:
- On the system that is being traced run blktrace in client
mode.
For example to trace the SCSI disk on device
/dev/sda and
send the trace data to the system t6345030. . . run:
# blktrace -h t6345030.mysystem.com -d /dev/sda
blktrace: connecting to t6345030.mysystem.com
blktrace: connected!
blktrace on the server side now shows
that there is a connection from the system being traced:
server: connection from 9.152.37.153
- Now run the I/O load that should be traced.
Afterward,
stop the blktrace client and then the blktrace server with Ctrl+C.
Both acknowledge this by printing a summary of the data that was traced:
Device: sda
CPU 0: 0 events, 66471 KiB data
CPU 1: 0 events, 53906 KiB data
Total: 0 events (dropped 0), 120377 KiB data
Results
The trace data is now available on the system where the
server side of blktrace was running:
# ls -l 9.152.37.153-2007-10-31-11\:38\:40/
total 120512
-rw-rw-r-- 1 schmichr schmichr 68065624 Oct 31 12:40 sda.blktrace.0
-rw-rw-r-- 1 schmichr schmichr 55199600 Oct 31 12:40 sda.blktrace.1