On 11/07/2013 12:58 PM, Dr.x wrote:
> *at 10:45 ,
>
> squid is very bad and cant login to ssh terminal and interrupts if terminal
> succeed*
>
> ==================================
> top command
> top - 14:07:43 up 1 day, 5:23, 4 users, load average: 0.11, 0.15, 0.15
> Tasks: 159 total, 1 running, 158 sleeping, 0 stopped, 0 zombie
> Cpu0 : 1.0%us, 0.3%sy, 0.0%ni, 93.9%id, 0.7%wa, 0.0%hi, 4.1%si, 0.0%st
> Cpu1 : 1.0%us, 1.0%sy, 0.0%ni, 97.2%id, 0.0%wa, 0.0%hi, 0.7%si, 0.0%st
> Cpu2 : 0.0%us, 0.0%sy, 0.0%ni, 99.7%id, 0.0%wa, 0.0%hi, 0.3%si, 0.0%st
> Cpu3 : 2.4%us, 1.7%sy, 0.0%ni, 94.8%id, 0.0%wa, 0.0%hi, 1.0%si, 0.0%st
> Cpu4 : 0.7%us, 0.7%sy, 0.0%ni, 97.3%id, 0.0%wa, 0.0%hi, 1.4%si, 0.0%st
> Cpu5 : 1.3%us, 0.7%sy, 0.0%ni, 96.3%id, 0.0%wa, 0.0%hi, 1.7%si, 0.0%st
> Cpu6 : 0.7%us, 0.7%sy, 0.0%ni, 97.0%id, 0.0%wa, 0.0%hi, 1.7%si, 0.0%st
> Cpu7 : 0.3%us, 0.0%sy, 0.0%ni, 99.7%id, 0.0%wa, 0.0%hi, 0.0%si, 0.0%st
> Mem: 8182336k total, 7957752k used, 224584k free, 172808k buffers
> Swap: 0k total, 0k used, 0k free, 6897624k cached
> ==================================================================
You did not post Squid processes as shown by top during this time :-(.
If you are having trouble capturing them, restrict top to showing Squid
effective user only (and sort by PID to avoid order changes). You can
also use top batch mode and then grep for results (but it is a little
trickier to configure top that way -- try 'W' or 'w' to save visual
configuration and read the manual page).
As of now, it is clear that your problems are not related to CPU and RAM
usage. The primary remaining suspects are:
* Squid (and other) processes blocked on disk I/O queues overflowing.
Top should show that, but you may need to watch through several top
screens to make sure none of the processes are in D or similar state.
* You are running out of some network-related resources such as
conntrack table, ephemeral ports, etc. See Amos email for more
information on that. To detect this, try checking system logs for kernel
and TCP stack errors. You may also want to count the number of TCP
connections (in all states) to see how many concurrent connections you
have to deal with and what states they are in.
> just to hint , the free memory before using squid is 1 g and after some
> time the free is about 200 M
> could this be the problem ???
Probably not. See my earlier explanation about "cached" file system buffers.
Also, while it is tempting to change random things to try to fix the
problem, please keep in mind that the more you change things, the higher
is the chance that the problem will move, invalidating all previous
discussions and results. That would be very confusing for everybody.
Identify the problem first. Then change one thing at a time to fix it.
Ideally, you should not use real users for testing and fixing this.
There are free tools that can reproduce most common HTTP proxying
environments in lab settings. (But then you would have to learn how to
use those tools correctly, of course).
HTH,
Alex.
Received on Thu Nov 07 2013 - 23:00:29 MST
This archive was generated by hypermail 2.2.0 : Fri Nov 08 2013 - 12:00:20 MST