Changes For This Release

• Multiple device support has been added. The if= and of= options now accept a comma (",") separated list of devices. Each device will be executed in its' own process, and if the procs= or slices= options are specified, additional processes will be created (as usual).

  Note: Multiple processes will become threads when the multi-threaded dt is ready.
  Also Note: Multiple devices on the command line is a short term solution, since the next major dt release will support a target input file to define many devices each with their own test options. But that said, the CLI method has been successful for 50+ devices on Windows (CLI length varies per host OS).

  So why add this support after all these years?
  well, for folks initiating their IO tools from NATE control servers, you'll realize the following benefits:

  - Reduces control server connections (e.g. rsh or ssh) to test clients.

  * Uses one connection for all devices, rather than one connection per device.

  - Reduces control server memory consumption when subtest()'ing with large objects.

  - dt does the scale out, so automation does not need to run on the test clients.

  So, what's different?

  - All process output is written to stdout, unless a log file is specified.

  - Each process' output is identified by the process ID (same as procs/slices).

  - By default, dt will wait for all device processes to complete before exiting.

  * For those desiring procs to terminate on error, please add "oncerr=abort".

  Please Note: To assure unique data per device, please consider adding the prefix= option. A standard prefix that has worked well is: prefix="%d@%h", where '%d' expands to the device name amd '%h' expands to the host name. This prefix is written at the start of each block, and has been instrumental in helping triage data corruptions.

• Updated log file format control strings to allow the device name "%dsf" or the full device path "%device" to be included in the log file name:

      Log File Format Keywords:
          %dsf   = The device name.       %device = The device path.
          %iodir = The I/O direction.     %iotype = The I/O type.
          %host  = The host name.         %pid    = The process ID.
          %user  = The user name.
  
        Example: log=dt_%host_%dsf-%iodir_%iotype-%pid.log
  

  So for each device, the device name/path is substituted into the log file name, to generate a unique log for each device. Note: Path delimiters: '/' or '\\', are converted to an understore '_'. For "%dsf", the device directory: '/dev/' or '\\.\' are stripped. See examples below.

  Please Note: Use of dt's log file option requires write access to the testbed directory, if you include a log path directory to this. On Windows, your testbed share must be mapped. On *nix, the auto-mounter takes care of this (of course), if client is properly configured.

• A new trigger type has been added, which uses the NetApp specific ZAPI API to send required commands to panic the cluster or specific nodes:

      Trigger Types:
          zapipanic = Panic controllers via ZAPI.
  

  At the present time, sending ZAPI commands is accomplished using this external tool:

  /usr/software/test/bin/ontapi on *nix systems.

  X:\\bin\\apitest\\apitest.exe on Windows systems.

  Later, ZAPI support may be integrated directly into dt (planned as part of threaded dt).

      NetApp Specific Options:
          ontap_cserver=string  The cluster server name.
          ontap_nodes=string    The cluster node name(s).
          ontap_username=string The user name (optional).
          ontap_password=string The password (optional).
          ontapi_path=string    The 'ontapi' path (optional)
  

• When attempting to execute the trigger script, if an unexpected exit status is returned, dt will now terminate instead of continuing to run. This unexpected exit status usually means the script path cannot be found, and a request from end users would prefer dt terminate, so it now will!

• On Windows, changed GetComputerName() to GetComputerNameEx() to avoid having the computer name truncated to 15 characters! This was evidenced in keepalive messages using "%h" or "%H", and when formatting log files with "%host".

• Fix bug when using enable=deleteperpass option. Was breaking out of the main test loop, if the destination directory previous existed, rather than being created by dt.

ssan-rhel5u4-21# cd /var/tmp
ssan-rhel5u4-21# mkdir dtlogs
ssan-rhel5u4-21# dt.last version
    --> Date: June 20th, 2012, Version: 17.66, Author: Robin T. Miller <--
ssan-rhel5u4-21# dt.last if=/dev/sdb,/dev/sdc,/dev/sdd,/dev/sde count=1 log=dtlogs/dt-%host-%dsf.log disable=compare,stats
ssan-rhel5u4-21# ls dtlogs
dt-ssan-rhel5u4-21-multi-dsfs.log  dt-ssan-rhel5u4-21-sdc.log-16709  dt-ssan-rhel5u4-21-sde.log-16711
dt-ssan-rhel5u4-21-sdb.log-16708   dt-ssan-rhel5u4-21-sdd.log-16710
ssan-rhel5u4-21# 
ssan-rhel5u4-21# rm dtlogs/*log*
ssan-rhel5u4-21# dt of=/var/tmp/dt.data1,/var/tmp/dt.data2 count=1 log=dtlogs/dt-%host-%device.log disable=stats enable=pdebug
ssan-rhel5u4-21# ls dtlogs
dt-ssan-rhel5u4-21-multi-devices.log            dt-ssan-rhel5u4-21-_var_tmp_dt_data2.log-22914
dt-ssan-rhel5u4-21-_var_tmp_dt_data1.log-22913
ssan-rhel5u4-21# 

X:\bin>x:\bin\dt.last if=\\.\PhysicalDrive1,\\.\PhysicalDrive2 log=c:\temp\dt-%host-%dsf.log count=1 disable=compare,stats

X:\bin>dir c:\temp\dt*log*
 Volume in drive C has no label.
 Volume Serial Number is 54F3-DB6A

 Directory of c:\temp

06/22/2012  10:11 AM               224 dt-rtmiller1-lxp-multi-dsfs.log
06/22/2012  10:11 AM               218 dt-rtmiller1-lxp-PhysicalDrive1.log
06/22/2012  10:11 AM               218 dt-rtmiller1-lxp-PhysicalDrive2.log
               3 File(s)            660 bytes
               0 Dir(s)  57,474,859,008 bytes free

X:\bin>

This first example is forcing a panic of all nodes via the cluster server (cserver).

Force an error condition that forces the trigger to be executed:

robin-ptc% ./dt if=/var/tmp/dt.data bs=512 count=1 position=0 \
                pattern=0xdeadface flags=direct \
                trigger=zapipanic ontap_cserver=rtmiller-vsim3-d2.sim.rtp.netapp.com

This second example forces a panic on specified cluster nodes:

robin-ptc% ./dt if=/var/tmp/dt.data bs=512 count=1 position=0 \
                pattern=0xdeadface flags=direct trigger=zapipanic \
                ontap_nodes=rtmiller-vsim3.sim.rtp.netapp.com,rtmiller-vsim4.sim.rtp.netapp.com

Last Modified: June 20th, 2012