Re: Process Resource Update

[Martin Pala]

Works well :) What about to add it to cvs?

It seems that monit reports the memory usage excluding cache related 
regions. The difference is quite significant (ca. 250MB)- i have 512MB RAM:

- monit reports ca. 135000 kB memory usage

- top, free and vmstat reports 382052 kB usage (of which ca. 135000 kB 
is related to programs and the rest is cache + buffers):

--8<--
unicorn:~/cvs/monit-proc# free
             total       used       free     shared    buffers     cached
Mem:        517268     379800     137468          0      15568     229016
-/+ buffers/cache:     135216     382052
--8<--

For memory sizing, it is important to know the real memory usage as 
monit reports now plus to know the buffer+cache portion too.

Thus - could be possible to report total memory usage with caches + 
buffers too? It could be probably good to use the "free's" output i.e. 
to report:
- total memory usage and
- memory usage without buffers and cache

Following statistic was catched later then above example (thus the 
numbers are different). It is possible to see that free, top and vmstat 
reports:

- MemTotal-MemFree ... as total memory usage
- MemTotal-MemFree-Buffers-Cached ... as memory usage without buffers 
and cache

unicorn:~/cvs/monit-proc# cat /proc/meminfo && free
(... memstat output:)
MemTotal:       517268 kB
MemFree:          5120 kB
Buffers:         26644 kB
Cached:         294828 kB
SwapCached:          0 kB
Active:         334888 kB
Inactive:       139624 kB
HighTotal:           0 kB
HighFree:            0 kB
LowTotal:       517268 kB
LowFree:          5120 kB
SwapTotal:     1052216 kB
SwapFree:      1052216 kB
Dirty:           49532 kB
Writeback:           0 kB
Mapped:         207584 kB
Slab:            29288 kB
Committed_AS:   361592 kB
PageTables:       1388 kB
VmallocTotal:   516020 kB
VmallocUsed:     21700 kB
VmallocChunk:   492896 kB

(... free output:)
             total       used       free     shared    buffers     cached
Mem:        517268     512148       5120          0      26644     294828
-/+ buffers/cache:     190676     326592
Swap:      1052216          0    1052216


Martin


Christian Hopp wrote:
> On Wed, 4 Aug 2004, Martin Pala wrote:
>
>
> > Christian Hopp wrote:
> >
> > > On Wed, 4 Aug 2004, Christian Hopp wrote:
> > >
> > >
> > >
> > > > Where ever I can, I can implement a sysdep version where I try to peek the
> > > > values directly from the system (like /proc/memory for linux).
> > >
> > >
> > > I have done an implementation for LINUX and SOLARIS, all other use the
> > > processtree code as a fallback for the total system memory.
> > >
> > > Christian
> >
> > I have tested it little bit again, works good :)
> >
> > I have added presentation of the values to the cervlet - see 'System'
> > row in the attached snapshot + extension patch (incremental)
>
>
> It is integrated now (:
>
>
> > Could be possible to add yet percentage of time spent in 'system'
> > (kernel space)? It seems that current CPU percentage is related to
> > userspace.
>
>
> Done.... user/system/(io)wait  (user for all, system+wait just linux and
> solaris)
>
>
> > Another value which could be useful is number of blocked processes
> > (which usualy means serious i/o problems). loadavg shows runnable
> > (ready) processes already - could be possible to add something like:
> >
> > loadavg_r[3] (formerly loadavg array)
> > loadavg_b[3]
>
>
> I need a some time to find out how to find it in the system.
>
> Christian
>
>
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> monit-dev mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/monit-dev