performance issue
Could anyone debug my mpstat output . Do i need to upgrade my cpu ?
Mpstat output
-
CPU minf mjf xcal intr ithr csw icsw migr smtx srw syscl usr sys wt idl
140 316 1063 654 10907138 1040 11301170 82
340 260 1248 816 10987638 1140 110710 130 77
CPU minf mjf xcal intr ithr csw icsw migr smtx srw syscl usr sys wt idl
1 2438 209 2639 1440 2212 600 110 3820 298431 522 16
3 2414 660 2771 1465 2842 739 110 3180 310334 541 10
CPU minf mjf xcal intr ithr csw icsw migr smtx srw syscl usr sys wt idl
10085 2491 1307 2245 451 201 4260 223923 430 33
360 175 2786 1556 3294 645 201 3600 278026 530 21
CPU minf mjf xcal intr ithr csw icsw migr smtx srw syscl usr sys wt idl
180 101 2469 1309 2285 504 215 4820 238026 450 29
330 178 2827 1603 3292 730 215 4140 296029 530 17
CPU minf mjf xcal intr ithr csw icsw migr smtx srw syscl usr sys wt idl
1110 116 2431 1309 2252 535 207 4860 239627 470 25
3230 154 2778 1604 3252 758 207 4370 311930 550 15
Pls advice
# 1
Please include more information...
Performance issue... when? Sometimes, all the time, every Sunday around 3 in the morning...
If I were a salesman, I'd surely say you need more CPUs, but that may not address your issue without further information.
Things to consider when writing a post -
1. People don't necessarily have logins to your system or know the history behind your machine.
2. Even though you know the issue inside and out, you still need to provide some history to the issue or question based on failure of requirement #1
3. When describing #2, please be specific as to any test cases or forensics you may have taken.
That being said, some questions are easy to ask - what are the arguments I need to run when doing a make on "splat", but some like please interpret my crash dump or core file or system performance, do in fact need more info.
Your initial post shows a cpu that is in the first set of data - please include the interval for mpstat you used - quite idle. In the following lines, you're incurring a fair amount of system activity, which could mean you're hitting swap or trying to open / close / read from a lot of files.
Just by chance, what does vmstat 10 show during the same test cycle?
Thank you for your patience,
Jeff
# 2
> Your initial post shows a cpu that is in the first set
> of data - please include the interval for mpstat you
> used - quite idle.
Actually, if the TS ran something like "vmstat 2" and then snarf&barf-ed the output straight onto the forum that first line would be the average values for each CPU over the _whole_ uptime of the system :) That's would explain the difference in idle times between the first and the rest of the lines.
But indeed: please include more info, so this doesn't tell us very much :)
If you're experiencing performance issues with your box then the questions Jeff posted are the first things you should be asking yourself. Once you've answered those you can start looking around using commands like mpstat, vmstat and iostat (among others).
Cheers!
Thomas
# 4
The answer is "most likely", however, not knowing whether sar is implemented on a system, not knowing the frequency or duration of each sampling, adhoc *stat commands can usually provide the same level of granularity and evidence to some system deficiency that properly configured sar can do.
Also, just in my own personal preference, I'm going to look at the *stat commands as first order of business when a problem is happening rather than wait for the next sar sampling to occur.
I'm wondering what the final outcome of this was since there has been no follow up reply to the first post....
Regards,
Jeff
