Hi, I need to find disk statistics for an environment that uses flexvols. I used to run "pvstats" written by Ben Swartzlander bswartz@netapp.combut that does not work in an flexvol environment.
I need something similar to that for be able to tell management where we stand.
I would appreciate to hear from you if you have something similar or if you perhaps modified Ben's code to work in an flexvol environment and you're willing to share.
Thanks, George
George Kahler email: george@yorku.ca Sr. Systems Administrator work: (416) 736-2100 x.22699 Computing and Network Services fax: (416) 736-5830 Ontario, Canada, M3J-1P3
------------------------------------------------------------------------------- There are three kinds of lies: lies, damned lies, and benchmarks - Mark Twain
Hi,
I can't find pvstats on NOW or Google. If it gives you reasonable data for traditional volumes, a (simple?) rewrite to monitor aggregates would do it. At the other hand you could run 'statit' on your storage controllers to see what your disks are doing, but you'll have to do the math by your own. (MS Excel will help you)
George Kahler schrieb:
Hi, I need to find disk statistics for an environment that uses flexvols. I used to run "pvstats" written by Ben Swartzlander bswartz@netapp.combut that does not work in an flexvol environment.
I need something similar to that for be able to tell management where we stand.
I would appreciate to hear from you if you have something similar or if you perhaps modified Ben's code to work in an flexvol environment and you're willing to share.
Two possible solutions (targets):
1. Look at qtree stats. 2. Look at the aggregate.
Poke around with this tool, it captures all of that: http://now.netapp.com/NOW/download/tools/perfstat
Disclaimer: The support center's ability to support anything performance related is limited to where you are right in the middle of an issue at the time.
Stetson M. Webster Managed Services Engineering NetApp Global Services - Southeast District
919.250.0052 Mobile Stetson.Webster@netapp.com www.netapp.com
-----Original Message----- From: Stefan Funke [mailto:bundy@usage.de] Sent: Friday, August 08, 2008 3:18 AM To: George Kahler Cc: toasters@mathworks.com Subject: Re: flexvol disk stats
Hi,
I can't find pvstats on NOW or Google. If it gives you reasonable data for traditional volumes, a (simple?) rewrite to monitor aggregates would do it. At the other hand you could run 'statit' on your storage controllers to see what your disks are doing, but you'll have to do the math by your own. (MS Excel will help you)
George Kahler schrieb:
Hi, I need to find disk statistics for an environment that uses flexvols. I used to run "pvstats" written by Ben Swartzlander bswartz@netapp.combut that does not work in an flexvol environment.
I need something similar to that for be able to tell management where we stand.
I would appreciate to hear from you if you have something similar or
if
you perhaps modified Ben's code to work in an flexvol environment and you're willing to share.
The one downfall of qtree stats is that it ONLY measures qtrees - not helpful for volumes without qtrees :(
I think some of this may be exposed via stats, but it's been a very long time since I've looked and I want to say that it wasn't there last time around.
-----Original Message----- From: owner-toasters@mathworks.com [mailto:owner-toasters@mathworks.com] On Behalf Of Webster, Stetson Sent: Friday, August 08, 2008 7:42 AM To: Stefan Funke; George Kahler Cc: toasters@mathworks.com Subject: RE: flexvol disk stats
Two possible solutions (targets):
1. Look at qtree stats. 2. Look at the aggregate.
Poke around with this tool, it captures all of that: http://now.netapp.com/NOW/download/tools/perfstat
Disclaimer: The support center's ability to support anything performance related is limited to where you are right in the middle of an issue at the time.
Stetson M. Webster Managed Services Engineering NetApp Global Services - Southeast District
919.250.0052 Mobile Stetson.Webster@netapp.com www.netapp.com
-----Original Message----- From: Stefan Funke [mailto:bundy@usage.de] Sent: Friday, August 08, 2008 3:18 AM To: George Kahler Cc: toasters@mathworks.com Subject: Re: flexvol disk stats
Hi,
I can't find pvstats on NOW or Google. If it gives you reasonable data for traditional volumes, a (simple?) rewrite to monitor aggregates would do it. At the other hand you could run 'statit' on your storage controllers to see what your disks are doing, but you'll have to do the math by your own. (MS Excel will help you)
George Kahler schrieb:
Hi, I need to find disk statistics for an environment that uses flexvols. I used to run "pvstats" written by Ben Swartzlander bswartz@netapp.combut that does not work in an flexvol environment.
I need something similar to that for be able to tell management where we stand.
I would appreciate to hear from you if you have something similar or
if
you perhaps modified Ben's code to work in an flexvol environment and you're willing to share.
I'm not sure what pvstats does or did but I've written a tool called 'topvol' which monitors volume activity and latency in a fashion similar to the top utility available on *nixes.
Topvol is a perl script that you run on a remote host and it connects to the filer and uses the stats command to collect volume ops and latency. To run topvol you'll need ssh key based access to the filer and a user account with privileges to run the stats command. Topvol then gathers the ops and latency statistics on all flexvols on the filer and sorts them in descending order, showing the top 20 volumes. You can have topvol sort by ops or latency. Here is an example usage and output:
sh-3.2# ./topvol filer 3 4 ops Ops Latency Volume ========================================= 427 3.51 melon_20K_fntst90 60 0.03 dmsapplog 41 0.98 vmware_its_pool_3_nfs 41 1.76 oak_20K_psoft_bin_n 35 6.83 redwood_20K_clone_psoft_bin_n 35 2.58 esx3b_jsl_nfs1 30 0.01 bb_garlic2_orasoft_nfs 25 41 ums_ms9 22 0.96 melon_20K_redoa 21 0.87 its_pool_4_nfs 21 0.01 msln_web_data 18 0.18 bb_garlic2_redoa 16 7.87 SWS_Exchange 15 18 ums_ms8 15 3.32 dmsosm1 13 13 msln_ms1_b 13 1.41 bb7prod_content_nfs 12 6.30 dms_caesar_dbdat0_2 10 0.45 dmsdbrlb_fc 8 35 melon_20K_cstyp2 8 0.20 melon_20K_psoft_bin
Above, I've given topvol the filer name, the interval I want it to collect results for, the number of iterations I want it to run and what I want it to sort by. In this case I had is sort by ops.
If you run topvol without any options it will give you a synopsis of what it is and how to run it.
While it does not collect historical results, you could redirect the output to a text file and let it run indefinetely then bring the results into excel and do some magic on it there.
You can find a copy here: http://communities.netapp.com/docs/DOC-1262
Hope it helps.
Romeo
On Thu, Aug 7, 2008 at 2:40 PM, George Kahler george@yorku.ca wrote:
Hi, I need to find disk statistics for an environment that uses flexvols. I used to run "pvstats" written by Ben Swartzlander bswartz@netapp.combut that does not work in an flexvol environment.
I need something similar to that for be able to tell management where we stand.
I would appreciate to hear from you if you have something similar or if you perhaps modified Ben's code to work in an flexvol environment and you're willing to share.
Thanks, George
George Kahler email: george@yorku.ca Sr. Systems Administrator work: (416) 736-2100 x.22699 Computing and Network Services fax: (416) 736-5830 Ontario, Canada, M3J-1P3
There are three kinds of lies: lies, damned lies, and benchmarks - Mark Twain
What are you actually trying to tell management?
I'd look at "stats show (volume|aggregate)" to get ops / second on your volume/aggregate, and if you actually want disk stats for a particular aggregate you could do a: "aggr status <agg name> -r" to get a list of disks and then do a "stats show disk".
filer1(takeover)> stats show aggregate aggregate:aggr0:total_transfers:0/s aggregate:aggr0:user_reads:0/s aggregate:aggr0:user_writes:0/s aggregate:aggr0:cp_reads:0/s aggregate:aggr0:user_read_blocks:0/s aggregate:aggr0:user_write_blocks:0/s aggregate:aggr0:cp_read_blocks:0/s
volume:testvol:avg_latency:0ms volume:testvol:total_ops:0/s volume:testvol:read_data:0b/s volume:testvol:read_latency:0ms volume:testvol:read_ops:0/s volume:testvol:write_data:0b/s volume:testvol:write_latency:0ms volume:testvol:write_ops:0/s volume:testvol:other_latency:0ms volume:testvol:other_ops:0/s
I¹d point to the ³average_latency² output as pretty management friendly information.
Also, MRTG, cricket, and toasterview all expose various pieces of information in graph form, and I don¹t know how it is where you work, but where I¹ve worked, management likes graphs better than spreadsheets. No slight against management folks, the whole point is to look at the big picture, right?
On 8/7/08 11:40 AM, "George Kahler" george@YorkU.CA wrote:
Hi,
I wrote a couple of scripts to get a quick view of a filer's overall volume/LUN/disk/... performance a while ago. They basically use rsh to fetch the output of the "stats" command on the filer, and display the results on a screen in a sort of realtime way.
Most colleagues and customers I give this to seem to like it, maybe some of you may find them useful. I put them up here:
Any feedback is appreciated!
Regards, Filip
On Sat, Aug 9, 2008 at 1:13 AM, Nicholas Bernstein nick@nicholasbernstein.com wrote:
What are you actually trying to tell management?
I'd look at "stats show (volume|aggregate)" to get ops / second on your volume/aggregate, and if you actually want disk stats for a particular aggregate you could do a: "aggr status <agg name> -r" to get a list of disks and then do a "stats show disk".
filer1(takeover)> stats show aggregate aggregate:aggr0:total_transfers:0/s aggregate:aggr0:user_reads:0/s aggregate:aggr0:user_writes:0/s aggregate:aggr0:cp_reads:0/s aggregate:aggr0:user_read_blocks:0/s aggregate:aggr0:user_write_blocks:0/s aggregate:aggr0:cp_read_blocks:0/s
volume:testvol:avg_latency:0ms volume:testvol:total_ops:0/s volume:testvol:read_data:0b/s volume:testvol:read_latency:0ms volume:testvol:read_ops:0/s volume:testvol:write_data:0b/s volume:testvol:write_latency:0ms volume:testvol:write_ops:0/s volume:testvol:other_latency:0ms volume:testvol:other_ops:0/s
I'd point to the "average_latency" output as pretty management friendly information.
Also, MRTG, cricket, and toasterview all expose various pieces of information in graph form, and I don't know how it is where you work, but where I've worked, management likes graphs better than spreadsheets. No slight against management folks, the whole point is to look at the big picture, right?
On 8/7/08 11:40 AM, "George Kahler" george@YorkU.CA wrote:
-- Nicholas Bernstein Technologist, Consultant, Instructor http://nicholasbernstein.com
Great idea! Have you considered using keyless ssh instead of rsh though? You should submit it to the communities site or the toolchest.
-Nick
On 8/13/08 6:08 AM, "Filip Sneppe" filip.sneppe@gmail.com wrote:
Hi,
I wrote a couple of scripts to get a quick view of a filer's overall volume/LUN/disk/... performance a while ago. They basically use rsh to fetch the output of the "stats" command on the filer, and display the results on a screen in a sort of realtime way.
Most colleagues and customers I give this to seem to like it, maybe some of you may find them useful. I put them up here:
Any feedback is appreciated!
Regards, Filip
On Sat, Aug 9, 2008 at 1:13 AM, Nicholas Bernstein nick@nicholasbernstein.com wrote:
What are you actually trying to tell management?
I'd look at "stats show (volume|aggregate)" to get ops / second on your volume/aggregate, and if you actually want disk stats for a particular aggregate you could do a: "aggr status <agg name> -r" to get a list of disks and then do a "stats show disk".
filer1(takeover)> stats show aggregate aggregate:aggr0:total_transfers:0/s aggregate:aggr0:user_reads:0/s aggregate:aggr0:user_writes:0/s aggregate:aggr0:cp_reads:0/s aggregate:aggr0:user_read_blocks:0/s aggregate:aggr0:user_write_blocks:0/s aggregate:aggr0:cp_read_blocks:0/s
volume:testvol:avg_latency:0ms volume:testvol:total_ops:0/s volume:testvol:read_data:0b/s volume:testvol:read_latency:0ms volume:testvol:read_ops:0/s volume:testvol:write_data:0b/s volume:testvol:write_latency:0ms volume:testvol:write_ops:0/s volume:testvol:other_latency:0ms volume:testvol:other_ops:0/s
I'd point to the "average_latency" output as pretty management friendly information.
Also, MRTG, cricket, and toasterview all expose various pieces of information in graph form, and I don't know how it is where you work, but where I've worked, management likes graphs better than spreadsheets. No slight against management folks, the whole point is to look at the big picture, right?
On 8/7/08 11:40 AM, "George Kahler" george@YorkU.CA wrote:
-- Nicholas Bernstein Technologist, Consultant, Instructor http://nicholasbernstein.com
-
Filip Sneppe -
George Kahler -
Glenn Walker -
Nicholas Bernstein -
Romeo Theriault -
Stefan Funke -
Webster, Stetson