Re: what is the command in PA to increase the default buffers

emanuel · ‎2009-10-23

Hello all

I need to increase Performance Advisor's ability to retain performance data from 5 days to 30 days or more on all default views. I heard it is a CLI option, where can I set that option?

Thanks, Emanuel

adaikkap · ‎2009-10-23

Will give the details of current retention.

[root@lnx ~]# dfm perf data list
Counter Group Host Name Sample Rate Retention Period Max Records Used Space Max Projected Space
------------------------- ---------------------------------------- --------------- ---------------- ---------- --------------- --------------------

Use dfm perf data modify to change the retention.

[root@lnx ~]# dfm perf datas modify help

NAME

modify -- Modify the interval and collection details for a counter group

SYNOPSIS

dfm perf data modify [ -f ] -G <counter-group-name> [ -o <host-name-or-id> ]

[ -s <sample-rate> ] [ -r <retention-period> ]

DESCRIPTION

Modify the interval and retention details for a counter group.

Specify a positive number, optionally followed by a time period

suffix to indicate seconds, minutes, hours, days, or weeks.

The default time period suffix is seconds.

'host-name-or-id' is mandatory for default counter groups.

Use -f option to force the change of sample rate and

retention period when the number of records for

the view decreases.

[root@lnx ~]#

Regards

adai

emanuel · ‎2009-10-23

Okay .. getting closer ... here is what I have found ( see below ) ... the 604800 is the 7 day collection period but I noticed that the used space is all different. I do not need to bump all these up but now i have new questions

Looking at the list of counter groups, I can see that I would want to increase "system basic", "disk basic", "aggregate basic", "NFS basic", "volume basic", and "NIC basic". I will be using the view in "View > Logical or Physical > controller > object" to dive down into a specific volume or disk to look at what the object was doing when a client side alarm goes off when it complains that our filers are slow. This is a strict NFS environment, no SAN or CIFS or iSCSI.

When I look at the collection period, I can do the math to bump up the coverage to 30 days ... if I do this on each group, does the "Max Records" increase as well with the "Used Space"? Is the Max Records something I can or need to modify? Is the Used Space a reflection of how large the data has grown since it has been collecting?

If I increase the collection period which will increase the size on disk for storing the data, is there a way I can size?

[root@sjcfilermon01 ~]# dfm perf data list

Counter Group Interval (secs) Collection Period (secs) Max Records Used Space (bytes)

------------------------- --------------- ------------------------- ---------- --------------------

System Basic 60 604800 10080 3709952

Disk Basic 60 604800 10080 1307174912

Aggregate Basic 60 604800 10080 18225152

CPU Basic 60 604800 10080 4032512

Network Interface Basic 60 604800 10080 28224512

NFS Basic 60 604800 10080 2419712

Target Basic 60 604800 10080 512

LUN Basic 60 604800 10080 512

Volume Basic 60 604800 10080 432392192

CIFS Basic 60 604800 10080 1129472

FCP Basic 60 604800 10080 1129472

iSCSI Basic 60 604800 10080 1129472

vFiler Basic 60 604800 10080 512

System Archive 60 31536000 525600 278225976

CPU Archive 60 31536000 525600 636930616

Perf Archive 600 31536000 52560 27364324

QOS Basic 60 604800 10080 512

Wafl Basic 60 604800 10080 4355072

Qtree Basic 60 604800 10080 512

Aggregate Health Chart 60 2592000 43200 7309880

Data Volume Latency 60 2592000 43200 22686288

SJC Trunked Interfaces 60 2592000 43200 4537772

Protocol Latency 60 2592000 43200 7058516

CATY16 NFS volume 60 2592000 43200 19526784

CATY18 NFS Volume 60 2592000 43200 17798784

System Level IO Profile 60 2592000 43200 18238112

adaikkap · ‎2009-10-23

Max records is=Collection Period/Sampling interval.

In your case =604800/60=10080. This will change automatically when you change your collection period or sampling.

If you reduce any of these that might reduce the max records then you must you -f options in the perf data modify.

-f it forces the change of sample rate and retention period when the number of records decrease

Used size is the currently used size.

Projected size is the max size to which this can grow with the current sampling and collection period. this field is available starting next release.

Regards

adai

emanuel · ‎2009-10-23

Okay, I have increased them to 30 day increments on the NAS/NFS specific counters ( ignoring iSCSI and FCP SAN counters ). I will check these next week to see if we keep longer than 7 days on default views.