Re: Performance of concurrent connections decreases from 1.5-dev7 to 1.5-dev17

Hi Godbach, On Sat, Jul 06, 2013 at 02:14:53PM +0800, Godbach wrote: I have tested dev17 again without the commit 6e0644339f3b and used MALLOC in pool_refill_alloc() instead of CALLOC. The result is nearly the same as that in dev7. Great, thank you for confirming. Again, you must keep in

Re: Performance of concurrent connections decreases from 1.5-dev7 to 1.5-dev17

On Sat, Jul 06, 2013 at 04:31:49PM +0800, Godbach wrote: I've got what you meant. Once allocated from system, haproxy will not release the memory until it receives SIGQUIT or does soft stop. If less memory during running is expected, just decrease the maxconn. In addition, can I consider

Re: Performance of concurrent connections decreases from 1.5-dev7 to 1.5-dev17

On 2013/7/6 16:52, Willy Tarreau wrote: On Sat, Jul 06, 2013 at 04:31:49PM +0800, Godbach wrote: I've got what you meant. Once allocated from system, haproxy will not release the memory until it receives SIGQUIT or does soft stop. If less memory during running is expected, just decrease the

Performance of concurrent connections decreases from 1.5-dev7 to 1.5-dev17

Hi Willy, I have done concurrent test for both haproxy 1.5-dev7 and 1.5-dev17.The amount of connections decreases in haproxy-1.5-dev17 compared to haproxy-1.5-dev7 under the the same condition and the same configuration file. I have gathered the memory usage by free command during test, and the

RE: Performance of concurrent connections decreases from 1.5-dev7 to 1.5-dev17

Hi! It seems that dev17 may use some additional memory allocated and freed besides pool. Any suggestions will be much appreciate. In my unscientific testcase I see increased memory consumption (RSS) since commit 6e0644339f3b (MEDIUM: memory: add the ability to poison memory at run time; its

Re: Performance of concurrent connections decreases from 1.5-dev7 to 1.5-dev17

Hi Lukas, hi Godbach, On Fri, Jul 05, 2013 at 09:27:01PM +0200, Lukas Tribus wrote: Hi! It seems that dev17 may use some additional memory allocated and freed besides pool. Any suggestions will be much appreciate. In my unscientific testcase I see increased memory consumption (RSS)

RE: Performance of concurrent connections decreases from 1.5-dev7 to 1.5-dev17

Hi! This is expected (when poisonning is enabled), because of the way buffers are allocated/released. It seems there is not much to disable here, the -dM startup argument just defines the actual (poisoning-)byte (defaulting to 0x50). So poisoning seems to be always on if I understand

Re: Performance of concurrent connections decreases from 1.5-dev7 to 1.5-dev17

On Fri, Jul 05, 2013 at 10:43:56PM +0200, Lukas Tribus wrote: Hi! This is expected (when poisonning is enabled), because of the way buffers are allocated/released. It seems there is not much to disable here, the -dM startup argument just defines the actual (poisoning-)byte (defaulting