On Thu Jan 11, 2001 at 09:40:40AM +0200, Igor Khasilev wrote:
> On Wed, 10 Jan 2001, Alexei A. Pogrebkov wrote:
>
> > Вот как выглядит смерть машины с oops'om 1.5.5 под BlackCat-6.2, glibc-2.1.3, ядро
>2.2.17
> > используется только как Web-акселератор:
> > (только что померла, решил написать - может кому поможет)
>
> Жаль, я не могу сказать ничего конкретного по этому поводу.
>
> Первым делом я попробовал-бы проапгрейдить glibc.
Я тоже использую BCL 6.2.
Glibc-2.1.3 и насколько я знаю это последняя для 6.2
Переход на 2.2 влечет за собой очень большие пермены.
> Вы всё-таки не показали свой конфиг - а было-бы полезно посмотреть.
Пжалуста.
Все что по дефолту и acl-и снес, чтобы лист не засорять.
Железки:
PII-300 на Asus P3BF 64Mb
Количество юзеров 10-15 мах в одно время
Сразу не падает (копит процессы видать гад) падает к коцу дня когда этого
никто не ожидает
##
# nameservers. Use your own, not our.
##
nameserver 192.168.1.1
nameserver 62.244.34.36
nameserver 62.244.31.1
nameserver 62.244.50.1
##
# Ports and address to use for HTTP and ICP
##
#bind ip_addr|hostname
http_port 3128
icp_port 3130
##
## Change euid to that user
##
## WARNING: if you use 'userid, then you 'reconfigure will not be able to
## open new sockets on reserved (< 1024) ports and will not be able
## to return to original userid.
##
userid squid
##
## Change root directory. If don't know exactly what are you doing -
## leave commented.
#chroot
##
# Logfile - just debug output
# When used in form 'filename [{N S}] [[un]buffered]'
# will be rotated automatically (up to N files up to S bytes in size)
##
logfile /var/log/oops/oops.log { 3 1m } unbuffered
##
# Accesslog - the same as for squid. Re rotating - see note for logfile
##
accesslog /var/log/oops/access.log
##
# Pidfile. for kill -1 `cat oops.pid` and for locking.
##
pidfile /usr/local/oops/logs/oops.pid
##
# Statistics file - once per minute flush some statistics to this file
##
statistics /var/log/oops/oops_statfile
##
# icons - where to find link.gif, dir.gif, binary.gif and so on (for
# ftp lists). If omitted - name of running host will be used. But
# using explicit names is better way.
##
#icons-host ss5.paco.net
#icons-port 80
#icons-path icons
##
# When total object volume in memory grow over this (this mean
# that cachable data from network came faster then we can save on disk)
# drop objects (without attempt to save on disk).
##
mem_max 24m
##
# Hint, how much cached objects keep in memory.
# When total amount become larger then this limit - start
# swaping cachable objects to disk
##
lo_mark 10m
##
# start random early drop when number of clients reach some level.
# this can protect you against attacks and against situation when
# oops cant handle too much connections. By default - 0 (or no limits).
##
start_red 20
##
# refuse any connection when number of already connected clients reach some
# level. By default - 0 (or no limits).
##
refuse_at 25
##
# if document contain no Expires: then expire after (in days)
# ftp-expire-value - expire time for ftp (in days)
##
default-expire-value 30
ftp-expire-value 30
##
# Maximum expite time - doc will not keep in cache more then
# this number of days (except if defaiult-expire-value used for this documeny)
##
max-expire-value 120
##
# in which proportion time passed since last document modification
# will accounted in expire time. For example, if last-modified-factor=5
# and there was passed 10 days since document modification, then expiration
# will be setted to 2 days in future (but no nore then max-expire-value)
##
last-modified-factor 7
##
# If you want not cache replies without Last-Modified:
# uncomment next line.
##
#dont_cache_without_last_modified
# run expire every ( in hours )
##
default-expire-interval 10
##
# icp_timeout - how long to wait icp reply from peer (in ms, e.g 1000 = 1sec)
##
icp_timeout 3600
##
# start disk cache cleanup when free space will be (in %%)
# As on the very large storages 1% is large space (1% from 9G is
# 90M), then on such storages you can set both disk-low-free and
# disk-ok-free to 0. Oops will start cleanup if it have less then 256
# free blocks(1M), and stop when it reach 512 bree blocks(2M).
##
disk-low-free 3
##
# stop disk cache cleanup when free space will be (in %%)
##
disk-ok-free 5
##
# Force_http11 - turn on http/1.1 for each request to document server
# This option required if module 'vary' used.
##
force_http11
##
# Always check document freshness, even it is not stale or expired
# This force Oops behave like squid - first check cached doc, then send
##
always_check_freshness
##
# If user-requestor aborted connection to proxy, but there was received more
# then some percent ot the document - then continue.
# default value - 75%
##
force_completion 85
##
# maximum size of the object we will cache
##
maxresident 3m
insert_x_forwarded_for yes
insert_via yes
##
# If host have several interfaces or aliases, use exactly
# this name when connecting to server:
##
connect-from proxy.lpet.lg.ua
##
# line 'parent ....' will force all connections (except to destinations
# in local-domain or local-networks) go through parent host
##
parent proxy2.is.com.ua 3128
##
# Never use "parent" when connecting to server in these domains
##
local-domain lpet.lan lpet.lg.ua
local-domain donass.uarc5.org is.net.ua is.com.ua piton.net
local-networks 192.168.10/24 62.244.34.36/32 62.244.35.96/29
##
# limit whole group to 8Kbytes per sec
##
# bandwidth 8k;
##
# limit request rate from this group (requests per second). This is crude,
# and must be used as last resort
##
# maxreqrate 100;
##
# Storage section
# Change this for your own situation. Oops can work without
# storages (using only in-memory cache).
##
##
# Storage description (can be several)
# path - filename of storage. can be raw device (be carefull!)
# size - size (of storage file). Can be smthng like 100k or 200m or 4g
# Size used only durig format process (oops -z).
##
storage {
path /var/spool/squid/oops_cache ;
# Size of the storage. Can be in bytes or 'auto'. Auto is
# usefull for pre-created storages or disk slices.
# NOTE: 'size auto' won't work for Linux on disk slices.
# To use large ( > 2G ) files run configure with --enable-large-files
size 300m ;
# You have to use 'offset' in the case your raw device (or slice)
# require that. For example if you use entire disk as storage
# under AIX and Soalris/Sparc - you have to skip first block
# which contain disk label (that is storage will start from
# next 512 sector.
# offset 512;
}
storage {
path /var/spool/squid/oops_cache1 ;
size 400m ;
}
--
Konstantin N. Bezruchenko
BK5536-RIPE
=====================================================================
If you would like to unsubscribe from this list send message to
[EMAIL PROTECTED] with "unsubscribe oops" in message body.
Archive is accessible on http://www.paco.net/oops/
