Subscribe
Highlighted

dfm data export versus NMC export

So, according to the documentation as well as my hours-on-end experimenting with this command, the CLI export functionality is completely inferior to the NMC in the following ways:

1) You can't specify a time range.  You can only specify how far back you go realative to the time the command is executed.

2) You can't specify a particular counter.  Instead, it pulles ALL data for the time specified.

3) You can't specify a particular filer or group when you schedule an export.  Instead, it pulls ALL data on ALL managed filers.

1 & 2 can be done via the NMC.

3 can't be done at all.

These limitations make scheduling reports on particular metrics very vexing.  Why can't the CLI have the functionality that the NMC has for 1 & 2 at least?  Unless the functionality is there but not documented, otherwise Netapp Development should really consider adding this ability.

Closely related to this is the "dfm perf view retrieve" command.  You can specify a time range there and particular filer or group, however, it pulles ALL the counters in a View.  Why can't Netapp just put in a flag to pull out a counter.

If anyone has any suggestions, I would greatly appreciate it.  Thanks.

Re: dfm data export versus NMC export

Hi,

<<<< 2) You can't specify a particular counter.  Instead, it pulles ALL data for the time specified.

You can add/delete counters exported via the 'dfm perf export counter add/delete' command.

dfm perf export counter list command displays the set of counters exported by default.

Check the help for more details.

<<<< 3) You can't specify a particular filer or group when you schedule an export.  Instead, it pulls ALL data on ALL managed filers.

Performance Advisor data export is controlled at the global level and at the per-host level  through perfDataExportEnabled option.

You can tweak this option to set this option only for the filers for which you want the data export to be enabled.

<<<< 1) You can't specify a time range.

I'll get back to you on this shortly.

Please refer the TR http://media.netapp.com/documents/tr-3690.pdf for more information and usage details of this feature

<<<< Why can't Netapp just put in a flag to pull out a counter.

The workaround for this functionality is to create a custom view with the counters that are of interest and use the 'dfm perf vierw retrieve' comman.

We appreciate your feedback on this feature + 'dfm perf view retrieve' command and shall consider this in our future releases.

Re: dfm data export versus NMC export

<<You can add/delete counters exported via the 'dfm perf export counter add/delete' command.

dfm perf export counter list command displays the set of counters exported by default.

Check the help for more details.>>

This functionality does not help me.  This command only adds or subtracts to what will be exported.  In other words, I would have to delete ALL the default exported counters that I did not want exported first, then run the export.  You still can't pick from the default list what counters you want to pull.  So, if I wanted to pull particular counters for one manager, then other counters for another manage, I would have to rebuild the default export list, which is very inefficient.  Also, the documentation is saying the "-v" option is to specify a "View."  This is missleading because a View is technically not what needs to be specified.  A View in PA is a collective grouping of selected counters taken from "counter groups".  What this option wants is a "counter object," i.e. the the object field in the fully qualified counter tuple: $filer:object:instance: counter:label:label, which is listed in the "dfm perf export counter list command."

<<Performance Advisor data export is controlled at the global level and at the per-host level  through perfDataExportEnabled option.

You can tweak this option to set this option only for the filers for which you want the data export to be enabled>>

Yes, this is true, however, this would mean that I would have to change the variables for different report request on different filers.  This would require alot of extra scripting work to automate.

Hope this makes sense.  Thanks alot for taking the time to respond!

Re: dfm data export versus NMC export

We have enhanced the 'dfm perf view retrieve' CLI in DFM 4.0 to just retrieve the data for a specific object instance and counter from a view.

Hope this would address most of your concerns.

The new help syntax from DFM 4.0 (not released yet) below for your information


[root@morpheus ~]# dfm perf view retrieve help

NAME
    retrieve -- retrieve the data associated with a view

SYNOPSIS
    dfm perf view retrieve [ -a <appliance-name-or-id> | -g <group-name-or-id> ]
                [ -o <perf-object> ] [ -i <perf-instance> ]
                <view-name> <starttime> <endtime>

DESCRIPTION
    retrieve the data associated with a view
        Format of starttime and endtime is "YYYY-MM-DD HH:MM:SS"

Re: dfm data export versus NMC export

The new CLI is "dfm perf data retrive" not "dfm perf view retrieve". The

syntax for the new cli is as follows:

  1. dfm perf data retrieve help

NAME

retrieve -- This command allows you to extract the counter data

with supported statistical calculations on them.

SYNOPSIS

dfm perf data retrieve { [ -C

perf-counter ... ]

}

[ -m statistical-method

]

DESCRIPTION

This command retrieves performance data for the specified counters and

object instances. These counters and instances can be specified

explicitly

using the -C and -o options or implicitly by specifying the view-name

using -V option.

When the view specified is associated with object types, the object

instance has to be explicitly specified using the -o option. When

the view

specified is associated with object instances, this option is not

mandatory. If -o option is specified in this case, the specified object

instances are considered, ignoring the instances in the view. If

counters are explicitly specified along with the view, the counters

in the view are ignored and the explicitly specified counters are

considered instead.

For finer control on the time, a filter with months, days and time

range with in a day can be specified. Only those time stamps that

satisfy the filter will be shown in the output. On the resulting

output of

counter data, statistical computations like minimum, maximum, mean,

and value_at_percentile can be computed.

Description of options:

-o Object instance for which data is to be retrieved. Multiple object

instances can be specified. Specifying a parent will retrieve data

for all its children.

-C A counter of the form object-name:counter-name (ex.

system:cpu_busy).

Multiple counters can be specified.

-V View Name. A view name can be specified, instead of specifying the

counters and instances explicitly.

-b Start time. Format: "yyyy-mm-dd hh:mm:ss". If this option is not

specified, the start time will be the time of the oldest record.

-e End time. Format: "yyyy-mm-dd hh:mm:ss". If this option is not

specified, the end time will be the time of the newest record.

-d Duration for which data is to be retrieved (in seconds). The time is

calculated going backwards from the current time. When this option

is specified, start-time and or end-time are ignored.

-s Sample rate (in seconds). This interval will be used to consolidate

the output data. The available data will be split into regions as

specified by the sample rate and the last sample in each of those

regions will be displayed. Also used for window calculation for

metrics.

-x Specifies the format to display the output. Possible values are

Legacy and TimeIndexed. Default value is Legacy.

-R If specified, data is rolled-up to the nearest minute. Applicable

only when the output format is TimeIndexed.

-M Filter based on month. Examples: Jan, Feb, etc. Comma separated

multiple values

can be specified.

-D Filter based on day of week. Examples: Mon, Tue, etc. Comma

separated multiple

values can be specified.

-T Filter based on time range in a day. Example: 16.00-21.55. Comma

separated multiple

values can be specified.

-m The statistical computation. Valid values are min, max, mean and

value_at_percentile.

-P For value_at_percentile type of calculation, this contains the

percentile value.

-S For computing on fixed size data, this defines the method to advance

the chunks of data. Valid values are simple, step and rolling.

Default

value is simple. Valid only when a statistical-method is specified.

Simple and step methods are available for all statistical methods.

Rolling is valid only for mean statistical method.