On 11/08/17 06:36, Cherukuri, Naresh wrote:

I cannot say by client or network. But, for sure I can say we have around 7000 
computers using squid as a proxy.

FYI: that number means peak load may be as high as 70K RPS (~100 req/client), a quad-core machine might be able to handle that but the 2.13 GHz CPU speed makes me doubtful. I'd plan for up to 4 machines of this type to be built in the medium-long term.

From: Cherukuri, Naresh
Sent: Thursday, August 10, 2017 16:27

No this a physical box and we are using only for squid. We have 4 cpu's and
16 cores. Please find below for reference Accesslogs : redirected to 

NP: I don't see any access logs details in your post. Just CPU specs and disk FS mappings.

-----Original Message-----
From: Cherukuri, Naresh
Sent: Thursday, August 10, 2017 16:03

Hello Eliezer,

We are using OS "Redhat 7" and squid version 3.5.20.

As of now we are not using any cache, we already commented out. You want me try 
using cache by uncommenting the following line.

# Uncomment and adjust the following to add a disk cache directory.
#cache_dir ufs /cache/squid 10000 16 256

No, suggestion was for the default memory-only cache. Which means no cache_mem or cache_dir entries in your squid.conf (letting Squid use its defaults). It should still be caching, just much less.

From: squid-users On Behalf Of Cherukuri, Naresh
Sent: Tuesday, August 8, 2017 16:28


I am new to squid. I am getting a problem every 19 hours squid takes all RAM 
memory, then started taking swap in  20 minutes my swap is full. Then server 
side (OOM) is activating and killing all squid child's then finally killing 
squid parent. Can someone help me how to address this problem?

FYI: My brief understanding of the OOM is that when the sum total of all processes on the machine start consuming too mush RAM it kills off the largest user. So it may be that Squid is using some large (but reasonable) amount of RAM and something else entirely pushes it over the edge - OOM just killing Squid because it has the most memory at that time.

So, if you have any record of the machines memory usage by process over time it would be good to know for certain whether it is Squid alone, or squid + something else that is the problem. Something along the lines of a log processor that kicks in once a day and uses lots of RAM briefly may exist.

Why every 19 hours my memory is going to full?

How much Ram do I need for following squid version?

Squid Cache: Version 3.5.20

              total       used       free     shared    buffers     cached
Mem:         11845       2713       9132         14         71       1641 -/+ 
buffers/cache:       1000      10845
Swap:        25551        421      25130

The biggest problem I see here is that there are no units. These could be KB or GB for all we know.

Please clarify that missing info.

FWIW: Squid with some minimal tuning runs fine on embeded devices with 32MB of RAM and has not had a memory leak in a long time. So it is usually a matter of some feature misconfigured. That said there is an issue with OpenSSL objects that can look like a memory leak in 3.x if one is not careful.

To see if anything is misconfigured please post your squid.conf (without the #commented out lines) so we can review it for problems.

squid-users mailing list

Reply via email to