Re: [cobalt-users] Question for everyone running 550 OS on XTR or even 550

[josh <josh@xxxxxxxxxxxxxxxxxx>]

I have 2GBs and it has occurred to me that it is also possible that
the 550 kernel does not handle 2G correctly, but I believe a genuine
RaQ550 will also take 2GB. I also thought the memory might be bad but
it only fails with the current kernel not the ROM kernel.

With squid I can swallow as much memory as I have. It will take 10M
per GIG of disk cache and then I can specify an amount of RAM to use
for caching as well. That will approximate the amount of memory squid
will use when I start the process - it will grow from there. The rough
formula is that is you want to use 150 MB you set up a 5GB disk cache
take 10M for each gig of cache space on disk, and then reserver 50MB
for use by users and over time another 50MB will grow into the mix. I
can add any number to cache_mem from within squid and squid will pull
it from the available memory when it restarts. My load isn't
great this machine. It has three functions; Web server for the Saratoga
Springs Public Library, Mail server for same, and Proxy server for our
public access computers. (I would be running a lot of perl scripts and
other things through web sites - using mysql etc. so the web server
keeps a lot of stuff going)

 There may be some other problems because there should be
processes running to reclaim memory from RAM. My guess is without
squid I wouldn't have noticed the problem until long after I assumed
it was all working. It is quite possible if the routine that reclaim
memory no longer in use from ram were working, and the swap was broken
I would not have reached the problem either.

  It was by taking snapshots with top every hour that I was able to
notice the two odd aspects. It is normal for swap to be used at some
point if you figured out your server. If it is never ever used you
have more server than you need. Ideally it should be rarely used.

So at 8:13 AM we were showing 

8:13am up 18:14, 1 user, load average: 0.00, 0.00, 0.00 129 processes:
126 sleeping, 1 running, 2 zombie, 0 stopped CPU states: 2.5% user,
1.7% system, 0.0% nice, 30.3% idle Mem: 2061956K av, 2027384K used,
34572K free, 0K shrd, 0K buff Swap: 524536K av, 0K used, 524536K free
1217076K cached

And then there was a push into swap right after we opened:

9:13am up 19:14, 1 user, load average: 0.08, 0.05, 0.01 137 processes:
133 sleeping, 1 running, 2 zombie, 1 stopped CPU states: 2.5% user,
1.7% system, 0.0% nice, 33.7% idle Mem: 2061956K av, 1994424K used,
67532K free, 0K shrd, 0K buff Swap: 524536K av, 144K used, 524392K
free 1222980K cached

As I write the cat /proc/meminfo yields this -- but in isolation and
without more information about memory in general it is hard to say
what is right and what is wrong.

Mem:  2111442944 2082205696 29237248        0        0 1637175296
Swap: 537124864 15740928 521383936
MemTotal:      2061956 kB
MemFree:         28552 kB
MemShared:           0 kB
Buffers:             0 kB
Cached:        1583432 kB
SwapCached:      15372 kB
Active:         424344 kB
Inactive:      1352052 kB
HighTotal:     1179648 kB
HighFree:         9992 kB
LowTotal:       882308 kB
LowFree:         18560 kB
SwapTotal:      524536 kB
SwapFree:       509164 kB

My load is very light. Squid when it has space to cache items in
memory and manage is maintaining trees (b+ ??) so that disk or ram
cache allows location and retrieval of cached web sites
instantaneously and it actually lightens the load on the server
because it has to grab fewer bytes from the net.

[16:38:43:war-admiral:/etc]uptime
  4:38pm  up 1 day,  2:37,  3 users,  load average: 0.02, 0.10, 0.05



On Fri, Dec 12, 2003 at 08:26:34AM -0600, Chris L wrote:



On Fri, Dec 12, 2003 at 08:26:34AM -0600, Chris L wrote:
> 
> > You have never used your swap! This is what I was afraid of seeing. I
> 
> Just out of curiosity, how much RAM do you have? Unless you have very little

2GB twice what you have.
-- 
Josh Kuperman                       
josh@xxxxxxxxxxxxxxxxxx