On 09/08/2015 10:39 AM, Jorgeley Junior wrote:
I have 8GB physical memory and my swap is 32GB. I didn't increase the swap yet, should I?
You must start with reading the memory FAQ: http://wiki.squid-cache.org/SquidFaq/SquidMemory The general rule for all processes applies: make sure that a process is *not* larger than 80% of the physical memory. In your case, you must reduce cache_mem and make sure that Squid does not use more than 6 GB. A swap of 32 GB is fine for a system with 8 GB physical memory. I also suggest to consider a memory upgrade. Marcus
2015-09-08 9:23 GMT-03:00 Marcus Kool <marcus.k...@urlfilterdb.com <mailto:marcus.k...@urlfilterdb.com>>: On 09/08/2015 08:11 AM, Jorgeley Junior wrote: Thank you all, this is the output: vm.overcommit_memory = 0 vm.swappiness = 60 I have a Redhat 6.6 The value of vm.overcommit_memory is OK. The default value for vm.swappiness is way too high. It means that Linux swaps out parts of processes when they are idle for a while. For better overall system performance, you want those processes in memory as long as possible and not swapped out so I recommend to change it to 15. This implies that the OS has 15% of the physical memory available for file system buffers which is plenty. You only mentioned that the swap is 32 GB. What is the size of the physical memory ? Did you already increase the swap ? Marcus 2015-09-05 15:08 GMT-03:00 Marcus Kool <marcus.k...@urlfilterdb.com <mailto:marcus.k...@urlfilterdb.com> <mailto:marcus.k...@urlfilterdb.com <mailto:marcus.k...@urlfilterdb.com>>>: On Linux, an important sysctl parameter that determines how Linux behaves with respect to VM allocation is vm.overcommit_memory (should be 0). And vm.swappiness is important to tune servers (should be 10-15). Which version of Linux do you have and what is the output of sysctl -a | grep -e vm.overcommit_memory -e vm.swappiness Marcus On 09/04/2015 07:04 PM, Jorgeley Junior wrote: Thanks Amos, i will increase the swap Em 04/09/2015 17:22, "Amos Jeffries" <squ...@treenet.co.nz <mailto:squ...@treenet.co.nz> <mailto:squ...@treenet.co.nz <mailto:squ...@treenet.co.nz>> <mailto:squ...@treenet.co.nz <mailto:squ...@treenet.co.nz> <mailto:squ...@treenet.co.nz <mailto:squ...@treenet.co.nz>>>> escreveu: On 5/09/2015 7:16 a.m., Jorgeley Junior wrote: > Thanks Amos, my swap is 32GB, so that's causing the error as you said. > Which is the better choice: increase the swap size or reduce the > cache_mem??? > Both probably. 128 GB swap I suspect you will need. Increase the swap so the system lets Squid use more virtual memory. Decrease the cache_mem so that Squid does not actually end up using the swap for its main worker processes. That is a real killer for performance. Amos _______________________________________________ squid-users mailing list squid-users@lists.squid-cache.org <mailto:squid-users@lists.squid-cache.org> <mailto:squid-users@lists.squid-cache.org <mailto:squid-users@lists.squid-cache.org>> http://lists.squid-cache.org/listinfo/squid-users -- *_ _* *_ _* _______________________________________________ squid-users mailing list squid-users@lists.squid-cache.org <mailto:squid-users@lists.squid-cache.org> http://lists.squid-cache.org/listinfo/squid-users -- *_ _* *_ _*
_______________________________________________ squid-users mailing list squid-users@lists.squid-cache.org http://lists.squid-cache.org/listinfo/squid-users
