Of showers and sysstat

I was asked by a client yesterday in passing how to check CPU utilization on one of their NetApp filers.  I didn’t immediately know where to go and we quickly moved on to something else.  So as I was showering this morning, as I do many mornings, I remembered that, of course, you can use sysstat to view performance data.  Anyways, this is a real nice way to view instantaneous general performance data.  Options for this command are shown below.

deathstar> sysstat ?
usage: sysstat [-c count] [-s] [-u | -x | -m | -f | -i | -b] [interval]
-c count        – the number of iterations to execute
-s              – print out summary statistics when done
-u              – print out utilization format instead
-x              – print out all fields (overrides -u)
-m              – print out multiprocessor statistics
-f              – print out FCP target statistics
-i              – print out iSCSI target statistics
-b              – print out SAN statistics
interval        – the interval between iterations in seconds, default is 15 seconds

So for a sample of utilization stats for 10 seconds, taken every second, we can use

deathstar> sysstat -i -c 10 -u 1
CPU   Total     Net   kB/s    Disk   kB/s    Tape   kB/s  Cache  Cache    CP  CP  Disk
          ops/s      in    out      read    write    read    write    age      hit       time  ty   util
  1%     234       0      0     156        24         0        0       >60     99%     0%     –     4%
  0%     107       1      0      28          0         0        0       >60    100%     0%     –     2%
  1%     223       0      0      36          0         0        0       >60    100%     0%     –     2%
  1%     169       0      0      80          0         0        0       >60     98%      0%     –     4%
  7%     247       1      0    3856   4592         0        0       >60    100%    20%    T    12%
  2%     156       0      5    1952  32248        0        0       >60     99%     97%     :    14%
  1%     207       0      0     168         0         0        0       >60     99%       0%     –     3%
  1%     252       1      0     128         0         0        0       >60     99%       0%     –     2%
  1%     231       0      0      96          0         0        0       >60     99%       0%     –     3%
  1%     173       0      0      84          0         0        0       >60    100%      0%     –     2%

On the far left we see CPU utilization, then total IOPS for all protocols, block and file and towards the right side we can see the Cache hit rate. This isn’t for Flash Cache, as I thought initially, but for memory cache hits.  Flash Cache performance data can be gathered using a preset Flash Cache performance counter, seen here

deathstar> stats list presets
Stats Presets:
  flexscale-access
  flexscale-pcs
  flexscale
  spcres_aggr
  spcres_vol

and

deathstar>stats show –p flexscale-access

This will spit out goodies such as cache usage %, hits/sec, misses/sec, and read and write chains.

An excellent article on performance monitoring of NetApp filers is right here…enjoy.

Google