Powershell Scripts for EMC Performance Statistics

Clint Kitson - EMC vSpecialist - @clintonskitson

EMC Community Network - Everything VMware at EMC
https://community.emc.com/community/connect/everything_vmware?view=overview

02/27/11 Version 1.0.1 (Limited testing so far, provided as-is and suggested currently for non-
production and lab use only, see MD5 checksums for the zip files at above URL)

TABLE OF CONTENTS
• Overview
• Powershell Scripts for Performance Statistics
o Celerra
o Clariion
o Symmetrix
• Outputting to a CSV file
• Leveraging Powershell and Collected Statistics
Overview

Introduction
This document is written to discuss and provide information around running powershell scripts for
collecting and viewing performance statistics on EMC storage arrays. Each script has default runtime
behavior that will return more generic results from each array. As well there are inputs that can be sent
to the script which will allow for the filtering of array objects. For example, if you have a NAS export
name, UUID, or name of a device the script will filter for relevant objects and return just those items. If
these items aren’t specified then most other items will be returned.

The scripts are written completely in powershell and leverage native commands from EMC management
utilities per array type to obtain the proper information. This means there is a dependency for each
script that the management software from the array you are trying to pull statistics from is running and
available and powershell is installed and properly allowing execution of scripts. See the details below
per script. The scripts themselves are not recording any statistics information. However, there is a
section below that gives an example of what extra commands to add at the end of powershell execution
to output to files and “other things.”

Performance stats are derived (script calculated) or calculated. The sampling time starts when the script
executes, so the minimum time to complete the script is based on the sampling interval.

Runtime Abilities
The scripts will by default output information in a standardized manner as a powershell object. This
allows for a lot of options for formatting and using the output in other ways. The runtime options for
each script can be simple-- which array to target and how to authenticate. From there you have
optional parameters to specify intervals, sleep times, and specific objects to query.

Security
I highly suggest that you create non-privileged monitoring accounts to access the management software
on each array. Symmetrix and Celerra monitoring utilizes SSH to pull information. The Clariion
leverages a local install of naviseccli and can use security files instead of specifying authentication.
However, commands are still placed in open scripts so it is possible to inject malicious code into
command statements. Please consider this and checking MD5 signatures of zip files before using.

Runtime Performance Notes

There is a high amount of processing that takes place on the powershell machine and some
parallelization that occurs when kicking off collections simultaneously. Make sure the system running
the powershell commands has available memory and CPU processing to ensure proper operations. On
the target side there are multiple collection commands executed at the same time. This can cause some
load on the target device, so it is important to ensure that the target array responds appropriately to
information gathering requests. Also make sure for the Symmetrix and Celerra runtime that your
reverse lookups are either working correctly or disabled via the SSH service. If not, there will be a 3-5
second delay for SSH sessions due to reverse lookups timing out.

Runtime options for all scripts

-username
-password
-paint = adding a -paint will tell the script to refresh stats to same window overwriting previous stats and
refresh those stats continuously until interrupted. A sleep parameter can be set for the VMax and
Celerra script.
-interval (15 seconds default) = the sampling period to collect performance stats
-metrics (default is different per script) = which metrics (columns) to be returned, * will return all
columns, separated by a comma with no space
-outcsv = output to a CSV file

Prerequisites
Powershell v2.0 http://support.microsoft.com/kb/968930
A “Set-executionpolicy unrestricted” must be run on the host before running these scripts unless you
sign them manually. See dependency sections per storage array type for other requirements.

Viewing from Powershell

When viewing the stats it is recommended that you resize your window and viewing area. You can do
this by going to properties on the powershell window and adjusting the sizes.

Powershell Scripts for Performance Statistics

EMC Celerra (NAS) Statistics – get_unified_nas_perf

Introduction
The get_unified_nas_perf script will return NAS performance statistics from a NS/VG/Celerra around
NFS, filesystems, volumes, cpu usage, and Ethernet network traffic.

Dependencies
The script requires the ssh_function.ps1 and sharssh folder with dotNet libraries in the same directory
as the script. The system that is running the script needs to be able to SSH to the control station of the
Celerra with the credentials that you specify.

Runtime parameters
-csip = Control Station IP

(optional)
-objects = server_stats -l lists all of the available objects that can be monitored
-nfsexport = The export name that is the focus of monitoring. With this command, there will be specific
volumes related to the export showed.
-datamover (default server_2) = Specify the datamover to query, if you specify nfsexport then this will
be ignored and determined automatically
-sleep (5 seconds default) = amount of time to wait in between set of performance samplings
-objects = Which performance objects should be queried, “server_stats server_2 -“ will list available
queries. This is not the same as –metrics which is purely a choice of which columns or metrics to be
shown.

Runtime Examples
(.\get_unified_nas_perf.ps1 -csip csip -username xx -password xx) – Return default performance
statistics for any object that has metrics for that time period.
(.\get_unified_nas_perf.ps1 -csip csip -username xx -password xx -objects
net.device,store,sys.cpuUtil,nfs.basic) – Return these specific performance stats (dart5 can be different
than dart6 on objects, for example sys.cpuUtil is different per)
(.\get_unified_nas_perf.ps1 -csip csip -username xx -password xx -nfsexport exportname -objects
net.device,store) – Return specific stats for a datamover and dependent volumes around a specific NFS
export.

Preflight Checks
SSH to the Control Station as the correct user. Run (export NAS_DB=/nas) if not logging in as nasadmin.
Run (/nas/bin/server_stats server_2). Live statistics should return.

Available Metrics
Name,Util %,Network In KiB/s,Network Out KiB/s,Network In Pkts/s,Network Out Pkts/s,Network In
Errors/s,Network Out Errors/s,Queue Depth,Read Ops/s,Write Ops/s,Read KiB/s,Write KiB/s,Avg Read
Size Bytes,Avg Write Size Bytes

Others stats are available, but the script is made mainly for these right now.

EMC Clariion (BLOCK) Statistics – get_unified_block_perf

Introduction
The get_unfiied_block_perf script will return block performance statistics from a NS/CX/Clariion array
around SP and LUN objects with metrics around utilization, IOs, bandwidth, cache, and FASTCache stats.
Traditional flare LUNs will return extensive statistics. Thin/thick luns will return statistics around the
virtual LUNs from the raidgroups of storage pools and not directly the thin/thick luns themselves. These
stats require an option in the runtime for engineering mode.

Dependencies
The script requires that Navisphere Server Software for Windows is installed. The script was tested
against “Windows Navisphere Server Software 7.30”. You can download from Powerlink ->
Support -> Software Downloads and Licensing -> Download J-O -> Navisphere Server Software. Copy the
naviseccli.exe executable after install from the Program Files (x86)-> EMC -> Navisphere CLI directory to
the directory where the scripts are being run from. In order for the array to being collecting statistics,
Statistics Gathering and Data logging must be enabled and started. Statistics can be enabled from
Unisphere -> Properties -> Statistics checkbox and Data logging can be enabled from Monitoring ->
Analyzer -> Performance Data Logging -> Start. Analyzer does not need to be licensed to use this script.

Runtime
-spaip = Storage Processor A management IP
-spbip = Storage Processor B management IP
-scope # = # is normally 0 for global, see naviseccli documentation

(optional)
-usesecfile = Use pre-determined authentication files. This can be run with “naviseccli -user admin -
password password -scope 0 -h spaip -addusersecurity”. Make sure to this twice, once for SPA and once
for SPB. With this specified -username and -password and -scope are not needed.
-eng = This switch must be followed by the engineering password for the Clariion. It will return more
metrics and objects (virtual luns from pools).
-lunid – Return only specific information based on this LUN

Runtime Examples
(.\get_naviseccli_perf.ps1 -spaip spaip -spbip spbip -username admin -password pass -scope 0)

Preflight Checks
After installing Navisphere Server Utility as described above and copied the naviseccli.exe file to the
script directory run (naviseccli -user xx -password xx -h spip -scope 0 getall -sp). It should return
information for that SP. Run the command on the second SP as well and make sure they both are
responding as expected. It is also a good idea if you are using a security file to test the above command
as (naviseccli -h spip getall -sp) for both SPs.

Available Metrics
EMC Symmetrix (BLOCK) Statistics - get_symclI_perf

Introduction
The get_sym_cli_perf command will return performance statistics from a Symmetrix array around
director, device, and disk objects with metrics around IO, bandwidth, and cache.
Device and grab metas and disks

Dependencies
The script executes commands via SSH to a system with symcli Solutions Enabler Linux installed and
working to the target array. The easiest method is to download the vApp and follow along with the
documentation. http://powerlink.emc.com -> Support -> Software Downloads and Licensing ->
Downloads S -> Solutions Enabler -> se___vapp.zip

Runtime
-symclihost – Target symcli host

(optional)
-lunid – Return only specific information based on this LUNID
-dev_name – return only specific devices making up the meta based on this primary device name
-sleep (5 seconds default) = amount of time to wait in between set of performance samplings

Runtime Examples
(.\get_symcli_perf.ps1 -symclihost symclihost -username user -password pass –paint) – Returns all
statistics that the stats command collecs data for in the default 15 second interval and paints them on
the screen in a continuous fashion.
Preflight Checks
After getting the solutions enabler and symcli working, SSH to the system that is running the symcli
software. From there execute (symstat -type REQUEST -i 5 -c 1). You should see a table with some
columns returned and if there is activity, stats returned as well.

Available Metrics
Name, r_per_second, w_per_second, io_per_second, kb_r_per_second, kb_w_per_second,
r_cache_hit_pct, w_cache_hit_pct, rw_cache_hit_pct, sequential_r_pct, wp_tracks

Outputting to a CSV file

Each script has the parameter of -outcsv where you can enter a csv file to output the performance
samplings to. These stats can then be compressed and sent via email, or you can leverage another script
to monitor the CSV files. More to come below on that..

Leveraging Powershell and Collected Statistics

Next