settrace

You can use the settrace command to set options to trace certain I/O operations through the system.

Syntax

Read syntax diagramSkip visual syntax diagram>>- svctask -- -- settrace -- -- -type --+- mdisk -+-- --------->
                                         '- vdisk -'

>-- -file -- filename_arg -- -- -trigger --+- full ----+-- ----->
                                           +- status --+
                                           +- command -+
                                           +- timeout -+
                                           +- trigger -+
                                           '- abort ---'

>--+----------+-- --+--------------+-- --+---------+-- --------->
   '- -abort -'     '- -timestamp -'     '- -data -'

>--+--------+-- --+-----------+-- --+---------+-- -------------->
   '- -tag -'     '- -detect -'     '- -init -'

>--+----------+-- --+---------+-- ------------------------------>
   '- -sense -'     '- -cmds -'

>--+--------------------------+-- ------------------------------>
   '- -percent -- percentage -'

>--+----------------------------------------------------+-- ---->
   '- -cmdlist -- cmd_list --+------------------------+-'
                             '- -cmdmask -- cmd_mask -'

>--+--------------------------------------------------------+--><
   '- -skcqlist -- skcq_list --+--------------------------+-'
                               '- -skcqmask -- skcq_mask -'

Parameters

-type mdisk | vdisk
Specifies the object type the options relate to.
-file filename_arg
Specifies the file name prefix for the trace file.
-trigger full | status | command | timeout | trigger | abort
Specifies the trigger option, that is, what to do when the trace is started (triggered).

The full, status, and command options are valid for both MDisks and VDisks. The timeout and trigger options are only valid for MDisks. The abort option is only valid for VDisks.

  • full = when trace buffer is full, stop, that is, do not wrap
  • status = when the given SCSI status (-skcqlist) is reported in sense data
  • command = when the given SCSI command (-cmdlist) is sent
  • timeout = when a timeout occurs
  • trigger = keep running until the trigger event, that is, wrap
  • abort = when an abort occurs
-abort
Optionally specifies the abort argument, which adds the abort details to the trace. This argument is only valid for VDisks.
-timestamp
Optionally specifies the time-stamp flag. Adds a time-stamp to each entry in the trace. A file name is created from the prefix plus a time-stamp. The file name is in the form <prefix>_NN_YYMMDD_HHMMSS, where NN is the current configuration node ID. Files are created in the /dumps/iotrace directory.
-data
Optionally specifies the data flag, which adds the I/O data to the trace.
-tag
Optionally specifies the ccb_tags flag. Adds the CCB tags to the trace. This argument is for valid for MDisks.
-detect
Optionally specifies the discovery flag. Adds the MDisk discovery details to the trace for MDisks.
-init
Optionally specifies the initialization flag, which adds the MDisk initialization details to the trace for MDisks.
-sense
Optionally specifies the sense flag, which adds the SCSI sense data to the trace. This flag is only valid for VDisks.
-cmds
Optionally specifies the commands flag, which adds the commands data to the trace. This flag is only valid for VDisks only.
-percent
Optionally specifies where in the trace file the chosen trigger point should occur. That is, this flag specifies how much data to gather after the trigger point. The default is 50%, so the trigger point would be in the middle of the trace file.
-cmdlist cmd_list
Optionally specifies a command list, which adds only those commands to the trace file.
-cmdmask cmd_mask
Optionally specifies a command mask, which adds only those commands to the trace file. This can only be entered if the -cmdlist argument has been entered.
-skcqlist skcq_list
Optionally specifies a SKCQ list, which adds only those SKCQ details to the trace file.
-skcqmask skcq_mask
Optionally specifies a SKCQ mask, which adds only those SKCQ details to the trace file. This can only be entered if the -skcqlist argument has been entered.

Description

This command sets the various I/O tracing options you want for a particular disk type, that is managed disks or virtual disks. When the relevant disk type trace is subsequently triggered, the options specify the data you want to include in the trace file.

The file name specifies a file name prefix for the trace file. The system appends the node panel name and a time-stamp to the file name. The node ID is the current configuration node.

A maximum of ten trace files will be kept on the cluster. When the eleventh trace is made, the oldest existing trace file is overwritten.

The directory may also hold files retrieved from other nodes. These files are not counted. SVC will delete the oldest file, when necessary, to maintain the maximum number of files.

Possible failures

Examples

An invocation example

svctask settrace -type vdisk -file tracedump -trigger abort
 -percent 100 -abort -timestamp