Summary: Cpu bottleneck ?

From: RaghuNath L(Raghu) <LRaghuNath_at_Lucent.Com>
Date: Wed Mar 26 2003 - 08:25:03 EST
Dear Managers,

Thanks to

Pablo jejcic,
Christophe Dupre,
Greg Ulyatt,
Tim Chipman,
Kevin Buterbaugh,
Darren Dunham ,
Morgon jones,
Cris Lovett,
Paulo Afonso Graner Fessel,
Hichael Morton,
John Stoffel.


Lot of people in the list suggested memory and I/O intensive application 
bottle necks but Christophe Dupre hit nail on it's head.

It was plain defragmentation of vxfs file system which solved the problem.

I can still see lot of I/O but not feeling any slowness in running 
Builds with this server.

Once agnin thanks All, Defrag method will availble in 
/opt/VRTSvxfsdoc/sys_admin direcotry
as pdf file.

Myself and Sun Asia pecific team had spent complete 7 day's and nights 
in one strech to find out any H/W problem.

Vxfs is extant based file systems requires defrag on file systems with 
high writes with some time line.

RaghuNath L
Systems Administrator
Lucent Technologies -IIDC
> 
> 
> -----Original Message-----
> From: sunmanagers-admin@sunmanagers.org
> [mailto:sunmanagers-admin@sunmanagers.org] On Behalf Of RaghuNath L(Raghu)
> Sent: 19 March 2003 10:09
> To: sunmanagers@sunmanagers.org
> Subject: Cpu bottleneck ?
> 
> 
> Dear Managers,
> 
> We facing performance problem with a 420 connected to T3+ (1 gb cache)this
> server used as Clearcase vob server.
> 
> Here are the statistics,
> 
> Top out put shows not much of user or kernel process running but still cpu
> availability is only 8 % but there is no cpu intensive applications running
> on it. It's a fully loaded 420R with 4 cpu 4 Gb Ram.
> 
> last pid: 14377;  load averages:  0.84,  0.72,  0.58
> 14:56:24
> 248 processes: 246 sleeping, 1 stopped, 1 on cpu
> CPU states:  8.0% idle,  5.2% user,  5.2% kernel, 81.6% iowait,  0.0% swap
> Memory: 4096M real, 62M free, 469M swap in use, 10G swap free
> 
> Mpstat output.
> 
> Out put shows processors are getting loaded ( data below idle column)
> 
> but there is no load by either user or kernel it's just around 11 % where
> all the cpu cycles are utilised ?
> 
> I/o wait is alarmingly high.
> 
> 
>      0    0   3 1201   194  193  225    1    9    5    0  1058    3   2
> 65  30
>      1    0   6  628   400  200  165    0   12    4    0   668    2   2
> 88   8
>      2    0   4  477   386  385  229    1   13    8    0   765    2   3
> 81  14
>      3    0   6  641   107  105  232    2    9    3    0   904    2   2
> 74  22
> CPU minf mjf xcal  intr ithr  csw icsw migr smtx  srw syscl  usr sys  wt idl
>      0    2  10 1682   200  198  180    3   17   10    0   914    3   3
> 94   0
>      1    2  14  569   404  200  298    3   14    3    0   591    1   1
> 89   9
>      2    1   4  328   439  439  234    1   12   17    0   708    1   1
> 54  44
>      3    1  10  745   112  109  246    1   24   15    0   611    2   2
> 90   6
> CPU minf mjf xcal  intr ithr  csw icsw migr smtx  srw syscl  usr sys  wt idl
>      0    5   4 1377   191  189  238    3   12    8    0   522    2   1
> 45  52
>      1    0   8  584   402  200  309    2   11    8    0   545    2   3
> 93   3
>      2    0   4  211   558  556  312    3   10    4    0  1055    3   2
> 63  32
>      3    0  12 1085   105  102  329    2   15   10    0   940    3   4
> 87   6
> CPU minf mjf xcal  intr ithr  csw icsw migr smtx  srw syscl  usr sys  wt idl
>      0    2   8 2082   187  186  193    2   20   10    0   948    1   2
> 73  24
>      1    0  14 1240   402  200  314    2    9    5    0   723    1   3
> 57  39
>      2    0  11  781   335  333  186    2   12   12    0   735    3   2
> 79  16
>      3   32   9 1868   103  101  178    2   17    7    0   498    0   2
> 97   1
> CPU minf mjf xcal  intr ithr  csw icsw migr smtx  srw syscl  usr sys  wt idl
>      0    1  17 1738   224  222  286    3   18    9    0  1767    3   3
> 52  42
>      1    0  15 1083   403  200  233    4   18   15    0  1171    3   3
> 79  15
>      2    0  20 1326   534  529  371    4   14   18    0  2291    8   5
> 77  11
>      3    0   9 1262   106  101  344    5   18   10    0  1861    5   4
> 91   0
> CPU minf mjf xcal  intr ithr  csw icsw migr smtx  srw syscl  usr sys  wt idl
>      0    1  30 3172   246  242  358    6   17   11    0  2010    3   6
> 66  25
>      1    1  17 1230   407  200  386    9   27   14    0  2271    8   5
> 72  15
>      2    2  26  823   570  562  280    9   21   13    0  1684    5   6
> 61  28
>      3    1  23 1570   109  102  363    7   21   10    0  1872    2   6
> 85   7
> CPU minf mjf xcal  intr ithr  csw icsw migr smtx  srw syscl  usr sys  wt idl
>      0    0  14 1764   226  223  348    5   27   14    0   838    2   4
> 80  13
>      1    1  19  940   403  200  321    3   37   20    0  1577    2   4
> 87   8
>      2    3  13 1498   570  557  350   13   36   19    0  1290    4   5
> 64  26
>      3    1  14  623   106  101  389    7   22   18    0  1581    3   7
> 74  16
> 
> But on sunfire 6800 cpu utilization is proportional to kernal and users
> processes .
> 
> load averages:  2.43,  2.30,  1.82
> 15:03:12
> 2871 processes:2857 sleeping, 2 zombie, 8 stopped, 4 on cpu
> CPU states: 67.9% idle, 20.3% user,  6.8% kernel,  5.0% iowait,  0.0% swap
> Memory: 16G real, 8315M free, 6718M swap in use, 14G swap free
> 
> Kidly help us i will summarize.
> 
> regards
> Raghu
> Ra
> _______________________________________________
> sunmanagers mailing list
> sunmanagers@sunmanagers.org
> http://www.sunmanagers.org/mailman/listinfo/sunmanagers
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers
Received on Wed Mar 26 08:29:15 2003

This archive was generated by hypermail 2.1.8 : Thu Mar 03 2016 - 06:43:07 EST