99% CPU usage in System (Was: Re: vinum in 4.x poor performer?)
Marc G. Fournier
scrappy at hub.org
Wed Feb 9 17:34:12 PST 2005
Most odd, there definitely has to be a problem with the Dual-Xeon ysystem
... doing the same vmstat on my other vinum based system, running more,
but on a Dual-PIII shows major idle time:
# vmstat 5
procs memory page disks faults cpu
r b w avm fre flt re pi po fr sr da0 da1 in sy cs us sy id
20 1 0 4088636 219556 1664 1 2 1 3058 217 0 0 856 7937 2186 51 15 34
20 1 0 4115372 224220 472 0 0 0 2066 0 0 35 496 2915 745 7 7 86
10 1 0 4125252 221788 916 0 0 0 2513 0 2 71 798 4821 1538 6 11 83
9 1 0 36508 228452 534 0 0 2 2187 0 0 46 554 3384 1027 3 8 89
11 1 0 27672 218828 623 0 6 0 2337 0 0 61 583 2607 679 3 9 88
16 1 0 5776 220540 989 0 0 0 2393 0 9 32 514 3247 1115 3 8 90
Which leads me further to believe this is a Dual-Xeon problem, and much
further away from believing it has anything to do with software RAID :(
On Wed, 9 Feb 2005, Marc G. Fournier wrote:
>
> still getting this:
>
> # vmstat 5
> procs memory page disks faults cpu
> r b w avm fre flt re pi po fr sr da0 da1 in sy cs us sy id
> 11 2 0 3020036 267944 505 2 1 1 680 62 0 0 515 4005 918 7 38 55
> 19 2 0 3004568 268672 242 0 0 0 277 0 0 3 338 2767 690 1 99 0
> 21 2 0 2999152 271240 135 0 0 0 306 0 6 9 363 1749 525 1 99 0
> 13 2 0 3001508 269692 87 0 0 0 24 0 3 3 302 1524 285 1 99 0
> 17 2 0 3025892 268612 98 0 1 0 66 0 5 6 312 1523 479 3 97 0
>
> Is there a way of determining what is sucking up so much Sys time? stuff
> like pperl scripts running and such would use 'user time', no? I've got some
> high CPU processes running, but would expect them to be shooting up the 'user
> time' ...
>
> USER PID %CPU %MEM VSZ RSS TT STAT STARTED TIME COMMAND
> setiathome 21338 16.3 0.2 7888 7408 ?? RJ 9:05PM 0:11.35
> /usr/bin/perl -wT /usr/local/majordomo/bin/mj_queuerun -v 0
> setiathome 21380 15.1 0.1 2988 2484 ?? RsJ 9:06PM 0:02.42
> /usr/bin/perl -wT /usr/local/majordomo/bin/mj_enqueue -r -d postgresql.org -l
> pgsql-sql -P10 -p10
> setiathome 21384 15.5 0.1 2988 2484 ?? RsJ 9:06PM 0:02.31
> /usr/bin/perl -wT /usr/local/majordomo/bin/mj_enqueue -r -d postgresql.org -l
> pgsql-docs -P10 -p10
> setiathome 21389 15.0 0.1 2720 2216 ?? RsJ 9:06PM 0:02.06
> /usr/bin/perl -wT /usr/local/majordomo/bin/mj_enqueue -r -d postgresql.org -l
> pgsql-hackers -P10 -p10
> setiathome 21386 13.7 0.1 2720 2216 ?? RsJ 9:06PM 0:02.03
> /usr/bin/perl -wT /usr/local/majordomo/bin/mj_enqueue -r -d postgresql.org -l
> pgsql-ports -P10 -p10
> setiathome 21387 13.2 0.1 2724 2220 ?? RsJ 9:06PM 0:01.92
> /usr/bin/perl -wT /usr/local/majordomo/bin/mj_enqueue -r -d postgresql.org -l
> pgsql-interfaces -P10 -p10
> setiathome 21390 14.6 0.1 2724 2216 ?? RsJ 9:06PM 0:01.93
> /usr/bin/perl -wT /usr/local/majordomo/bin/mj_enqueue -o -d postgresql.org -l
> pgsql-performance -P10 -p10
> setiathome 21330 12.0 0.2 8492 7852 ?? RJ 9:05PM 0:15.55
> /usr/bin/perl -wT /dev/fd/3//usr/local/www/mj/mj_wwwusr (perl5.8.5)
> setiathome 7864 8.9 0.2 8912 8452 ?? RJ 7:20PM 29:54.88
> /usr/bin/perl -wT /usr/local/majordomo/bin/mj_trigger -t hourly
>
> Is there some way of finding out where all the Sys Time is being used?
> Something more fine grained them what vmstat/top shows?
>
>
> On Wed, 9 Feb 2005, Loren M. Lang wrote:
>
>> On Wed, Feb 09, 2005 at 02:32:30AM -0400, Marc G. Fournier wrote:
>>>
>>> Is there a command that I can run that provide me the syscall/sec value,
>>> that I could use in a script? I know vmstat reports it, but is there an
>>> easier way the having to parse the output? a perl module maybe, that
>>> already does it?
>>
>> vmstat shouldn't be too hard to parse, try the following:
>>
>> vmstat|tail -1|awk '{print $15;}'
>>
>> To print out the 15th field of vmstat. Now if you want vmstat to keep
>> running every five seconds or something, it's a little more complicated:
>>
>> vmstat 5|grep -v 'procs\|avm'|awk '{print $15;}'
>>
>>>
>>> Thanks ...
>>>
>>> On Wed, 9 Feb 2005, Marc G. Fournier wrote:
>>>
>>>> On Tue, 8 Feb 2005, Dan Nelson wrote:
>>>>
>>>>> Details on the array's performance, I think. Software RAID5 will
>>>>> definitely have poor write performance (logging disks solve that
>>>>> problem but vinum doesn't do that), but should have excellent read
>>>>> rates. From this output, however:
>>>>>
>>>>>> systat -v output help:
>>>>>> 4 users Load 4.64 5.58 5.77
>>>>>
>>>>>> Proc:r p d s w Csw Trp Sys Int Sof Flt
>>>>>> 24 9282 949 8414***** 678 349 8198
>>>>>
>>>>>> 54.6%Sys 0.2%Intr 45.2%User 0.0%Nice 0.0%Idl
>>>>>
>>>>>> Disks da0 da1 da2 da3 da4 pass0 pass1
>>>>>> KB/t 5.32 9.50 12.52 16.00 9.00 0.00 0.00
>>>>>> tps 23 2 4 3 1 0 0
>>>>>> MB/s 0.12 0.01 0.05 0.04 0.01 0.00 0.00
>>>>>> % busy 3 1 1 1 0 0 0
>>>>>
>>>>> , it looks like your disks aren't being touched at all. You are doing
>>>>> over 99999 syscalls/second, though, which is mighty high. The 50% Sys
>>>>> doesn't look good either. You may have a runaway process doing some
>>>>> syscall over and over. If this is not an MPSAFE syscall (see
>>>>> /sys/kern/syscalls.master ), it will also prevent other processes from
>>>>> making non-MPSAFE syscalls, and in 4.x that's most of them.
>>>>
>>>> Wow, that actually pointed me in the right direction, I think ... I just
>>>> killed an http process that was using alot of CPU, and syscalls drop'd
>>>> down to a numeric value again ... I'm still curious as to why this only
>>>> seem sto affect my Dual-Xeon box though :(
>>>>
>>>> Thanks ...
>>>>
>>>> ----
>>>> Marc G. Fournier Hub.Org Networking Services
>>>> (http://www.hub.org)
>>>> Email: scrappy at hub.org Yahoo!: yscrappy ICQ:
>>>> 7615664
>>>> _______________________________________________
>>>> freebsd-questions at freebsd.org mailing list
>>>> http://lists.freebsd.org/mailman/listinfo/freebsd-questions
>>>> To unsubscribe, send any mail to
>>>> "freebsd-questions-unsubscribe at freebsd.org"
>>>>
>>>
>>> ----
>>> Marc G. Fournier Hub.Org Networking Services
>>> (http://www.hub.org)
>>> Email: scrappy at hub.org Yahoo!: yscrappy ICQ:
>>> 7615664
>>> _______________________________________________
>>> freebsd-questions at freebsd.org mailing list
>>> http://lists.freebsd.org/mailman/listinfo/freebsd-questions
>>> To unsubscribe, send any mail to
>>> "freebsd-questions-unsubscribe at freebsd.org"
>>
>> --
>> I sense much NT in you.
>> NT leads to Bluescreen.
>> Bluescreen leads to downtime.
>> Downtime leads to suffering.
>> NT is the path to the darkside.
>> Powerful Unix is.
>>
>> Public Key: ftp://ftp.tallye.com/pub/lorenl_pubkey.asc
>> Fingerprint: B3B9 D669 69C9 09EC 1BCD 835A FAF3 7A46 E4A3 280C
>>
>>
>
> ----
> Marc G. Fournier Hub.Org Networking Services (http://www.hub.org)
> Email: scrappy at hub.org Yahoo!: yscrappy ICQ: 7615664
> _______________________________________________
> freebsd-questions at freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-questions
> To unsubscribe, send any mail to "freebsd-questions-unsubscribe at freebsd.org"
>
----
Marc G. Fournier Hub.Org Networking Services (http://www.hub.org)
Email: scrappy at hub.org Yahoo!: yscrappy ICQ: 7615664
More information about the freebsd-questions
mailing list