Re: [mv] Re: MV 4.04a - Memory usage

[Victor Nolton <ven@pragakhan.com>]

>Quoting Jim Toro (jimtoro@hoflink.com):
>>  tatus: RO
>>  Content-Length: 1638
>>  Lines: 44
>>
>>
>>  I have been working on a simple database of products for one of my
>>  customers and the database is only about 800 items.  However when the
>>  following search is processed.
>>
>>     [page search="
>>  	sf=category
>>  	tf=title
>>  	se=[cat-data cat selector]
>>  	ml=7
>>  	va=perf_group=[cat-data cat selector]
>>  	"
>>         ]<center><IMG SRC="[cat-data cat linkimage]" ALT="[cat-data cat
>>  name]" BORDER=0></center>[/page]
>>
>>  It will get an internal server error if I allow the ML=  line to be above
>>  7.  When the search is done there are 490 out of the total of 800 items
>>  found in this search.
>>
>>  When it was alot less hits (say 200) I could have the ML= set up to
>>  about 20.  When the hits  got to around 400 it had to be brought down
>>  to ML=11 or so.  Now its got to be at ML=7.  As I add more items the ML=
>>  needs to be lower and lower but it seems this is dependant on the hits
>>  found. Any other search where the hits are low, I can set the ML much
>>  higher.
>>
>>  Anyone know why this would be happening.  MV seems to eat tons of ram when
>>  it is doing this.  Is there something I can edit in catalog.cfg or
>>  something else to make it use less ram.  I cant figure out why it would be
>>  eating this much ram and dying PLUS why the ML= must be set so low.
>>
>>  I am running MV 4.04a, BSDI 3.1, Apache 1.3.12 on a Pentium 266 with
>>  128megs ram.  System runs flawlessly with everything else.
>
>This is the classic BSD mystery problem. I have no idea why this
>happens, but it has affected every older system with borderline RAM that
>I have seen on BSD, in versions of Minivend from 2.03 all the way
>through Interchange 4.6.x.
>
>I believe it is related to signals and Perl, but I cannot be sure. If
>you run in a debugging environment, you will probably see an "out of
>memory!" error. In any case, I am betting it affects everything prior to
>Interchange 4.7.x on BSD unless you have plenty of RAM, and sometimes
>even then on a busy system.


I do know that if a process takes tooo long you get a internal server 
error. it's all cpu, ram and apache specific I think.

for instance. We have a old perl based mailing list.
1 is on a AMD 800 with 512 megs of ram
2 is on a amd 450 with 256 megs of ram

when i send a message out to people on server 1 it processes them and 
then later I get the screen saying it's done.

on server 2 with just about the same amount of users, I get a 
internal server error. BUT the server still processes all emails. it 
just takes server 2 tooo long to respond to say, HEY I'm done! it 
just takes to long apache gets annoyed and instead of waiting a few 
more minutes it dumps the error instead of a spash screen when 
completed.

so in this case, it could be it takes the system just a little TOO 
long to find the extra 4 and by that time apache says, something 
broke.. internal server error. this is a... unofficial theory :)

I was going to experiment with this. DB in memory setup
Database  products MEMORY 1
but haven't tried it yet. Not sure if it would help speed things up.

_______________________________________________
Minivend-users mailing list
Minivend-users@lists.akopia.com
http://lists.akopia.com/mailman/listinfo/minivend-users