Re: weird problem
From: Jarek Poplawski <hidden>
Date: 2009-07-14 16:24:51
On Tue, Jul 14, 2009 at 01:26:46AM +0200, Paweł Staszewski wrote:
Jarek Poplawski pisze:quoted
On Fri, Jul 10, 2009 at 04:47:54PM +0200, Jarek Poplawski wrote:quoted
On Fri, Jul 10, 2009 at 01:59:00AM +0200, Paweł Staszewski wrote:quoted
Today i make other tests with change of /proc/sys/net/ipv4/rt_cache_rebuild_count and kernel 2.6.30.1 And when rt_cache_rebuild_count is set to "-1" i have always load on x86_64 machine approx 40-50% of each cpu where network card is binded by irq_aff when rt_cache_rebuild_count is set to more than "-1" i have 15 to 20 sec of 1 to 3% cpu and after 40-50% cpu... Here is one more patch for testing (with caution!). It adds possibility to turn off cache disabling (so it should even more resemble 2.6.28) after setting: rt_cache_rebuild_count = 0 I'd like you to try this patch: 1) together with the previous patch and "rt_cache_rebuild_count = 0" to check if there is still the difference wrt. 2.6.28; Btw., let me know which /proc/sys/net/ipv4/route/* settings do you need to change and why 2) alone (without the previous patch) and "rt_cache_rebuild_count = 0" 3) if it's possible to try 2.6.30.1 without these patches, but with default /proc/sys/net/ipv4/route/* settings, and higher rt_cache_rebuild_count, e.g. 100; I'm interested if/how long it takes to trigger higher cpu load and the warning "... rebuilds is over limit, route caching disabled"; (Btw., I wonder why you didn't mention about these or maybe also other route caching warnings?)Here is take 2 to respect setting "rt_cache_rebuild_count = 0" even after cache rebuild counter has been increased earlier. (Btw, don't forget about this setting after going back to vanilla kernel.)Applied to 2.6.30.1 1) With rt_cache_rebuild_count = 0 grep . /proc/sys/net/ipv4/route/* /proc/sys/net/ipv4/route/error_burst:1250 /proc/sys/net/ipv4/route/error_cost:250 /proc/sys/net/ipv4/route/gc_elasticity:4 /proc/sys/net/ipv4/route/gc_interval:15 /proc/sys/net/ipv4/route/gc_min_interval:0 /proc/sys/net/ipv4/route/gc_min_interval_ms:0 /proc/sys/net/ipv4/route/gc_thresh:190536 /proc/sys/net/ipv4/route/gc_timeout:15 /proc/sys/net/ipv4/route/max_size:1524288 /proc/sys/net/ipv4/route/min_adv_mss:256 /proc/sys/net/ipv4/route/min_pmtu:552 /proc/sys/net/ipv4/route/mtu_expires:600 /proc/sys/net/ipv4/route/redirect_load:5 /proc/sys/net/ipv4/route/redirect_number:9 /proc/sys/net/ipv4/route/redirect_silence:5120 /proc/sys/net/ipv4/route/secret_interval:3600 I tune this route parameters after looking of traffic/route cache to have not many entries in cache that are not needed anymore so gc_timeout = 15 limit of max entries = 1524288 And make route cahce a little more "faster" for me after tune gc_elasticity secret_interval gc_interval gc_thresh So with this parameters 15 sec of something like this: 00:41:23 CPU %usr %nice %sys %iowait %irq %soft %steal %guest %idle 00:41:24 all 0.00 0.00 0.12 0.00 1.49 10.46 0.00 0.00 87.92 00:41:24 0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 00:41:24 1 0.00 0.00 0.00 0.00 4.00 36.00 0.00 0.00 60.00 00:41:24 2 0.00 0.00 0.00 0.00 8.91 47.52 0.00 0.00 43.56 00:41:24 3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 00:41:24 4 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 00:41:24 5 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 00:41:24 6 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 00:41:24 7 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 and 15 sec of something like this: 00:41:44 CPU %usr %nice %sys %iowait %irq %soft %steal %guest %idle 00:41:45 all 0.00 0.00 0.00 0.00 0.00 0.42 0.00 0.00 99.58 00:41:45 0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 00:41:45 1 0.00 0.00 0.00 0.00 0.00 1.00 0.00 0.00 99.00 00:41:45 2 0.00 0.00 0.00 0.00 0.00 2.04 0.00 0.00 97.96 00:41:45 3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 00:41:45 4 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 00:41:45 5 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 00:41:45 6 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 00:41:45 7 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 So i change /proc/sys/net/ipv4/route/gc_timeout to 1 with rt_cache_rebuild_count = 0 And output is like 20 sec of something like this 00:48:52 CPU %usr %nice %sys %iowait %irq %soft %steal %guest %idle 00:48:53 all 0.00 0.00 0.19 0.00 0.19 0.58 0.00 0.00 99.03 00:48:53 0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 00:48:53 1 0.00 0.00 0.99 0.00 0.99 0.00 0.00 0.00 98.02 00:48:53 2 0.00 0.00 0.00 0.00 0.00 2.00 0.00 0.00 98.00 00:48:53 3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 00:48:53 4 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 00:48:53 5 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 00:48:53 6 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 00:48:53 7 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 and after this two second of something like this: 00:48:49 CPU %usr %nice %sys %iowait %irq %soft %steal %guest %idle 00:48:50 all 0.00 0.00 0.09 0.00 0.27 2.17 0.00 0.00 97.46 00:48:50 0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 00:48:50 1 0.00 0.00 0.00 0.00 1.96 6.86 0.00 0.00 91.18 00:48:50 2 0.00 0.00 0.00 0.00 0.99 16.83 0.00 0.00 82.18 00:48:50 3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 00:48:50 4 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 00:48:50 5 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 00:48:50 6 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 00:48:50 7 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 00:48:50 CPU %usr %nice %sys %iowait %irq %soft %steal %guest %idle 00:48:51 all 0.00 0.00 0.00 0.00 1.86 10.41 0.00 0.00 87.73 00:48:51 0 0.00 0.00 0.00 0.00 0.00 1.00 0.00 0.00 99.00 00:48:51 1 0.00 0.00 0.00 0.00 4.85 26.21 0.00 0.00 68.93 00:48:51 2 0.00 0.00 1.00 0.00 5.00 29.00 0.00 0.00 65.00 00:48:51 3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 00:48:51 4 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 00:48:51 5 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 00:48:51 6 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 00:48:51 7 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00
Could you remind us how it differs from 2.6.28 with the same settings?
Another test: gc_timeout = 1 rt_cache_rebuild_count = 100 10 to 14 sec of something like this: 00:51:36 CPU %usr %nice %sys %iowait %irq %soft %steal %guest %idle 00:51:37 all 0.00 0.00 0.00 0.00 0.00 0.27 0.00 0.00 99.73 00:51:37 0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 00:51:37 1 0.00 0.00 0.00 0.00 0.00 2.00 0.00 0.00 98.00 00:51:37 2 0.00 0.00 0.00 0.00 0.00 1.00 0.00 0.00 99.00 00:51:37 3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 00:51:37 4 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 00:51:37 5 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 00:51:37 6 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 00:51:37 7 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 and two seconds of 10 to 30% cpu load more 2). Only last patch and almost all the time output like this 00:59:49 CPU %usr %nice %sys %iowait %irq %soft %steal %guest %idle 00:59:50 all 0.00 0.00 0.13 0.00 1.73 8.00 0.00 0.00 90.13 00:59:50 0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 00:59:50 1 0.00 0.00 0.00 0.00 4.00 24.00 0.00 0.00 72.00 00:59:50 2 0.00 0.00 0.00 0.00 8.91 34.65 0.00 0.00 56.44 00:59:50 3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 00:59:50 4 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 00:59:50 5 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 00:59:50 6 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 00:59:50 7 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 sometimes after 15 to 30 sec i have 1 to 2% cpu load
And how long do you have this 1 to 2% load? Is it with: rt_cache_rebuild_count = 0 gc_timeout = 1? Maybe you could describe the main difference with or without the first patch?
3). with default settings and without this patch i have almost all the time output like this:
You mean without these two patches, right? So, there is no breaks with less load like above?
01:21:40 CPU %usr %nice %sys %iowait %irq %soft %steal %guest %idle 01:21:41 all 0.00 0.00 0.00 0.00 2.14 10.97 0.00 0.00 86.89 01:21:41 0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 01:21:41 1 0.00 0.00 0.00 0.00 6.93 34.65 0.00 0.00 58.42 01:21:41 2 0.00 0.00 0.00 0.00 7.07 42.42 0.00 0.00 50.51 01:21:41 3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 01:21:41 4 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 01:21:41 5 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 01:21:41 6 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 01:21:41 7 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 with my settings: /proc/sys/net/ipv4/route/error_burst:1250 /proc/sys/net/ipv4/route/error_cost:250 /proc/sys/net/ipv4/route/gc_elasticity:4 /proc/sys/net/ipv4/route/gc_interval:15 /proc/sys/net/ipv4/route/gc_min_interval:0 /proc/sys/net/ipv4/route/gc_min_interval_ms:0 /proc/sys/net/ipv4/route/gc_thresh:190536 /proc/sys/net/ipv4/route/gc_timeout:15 /proc/sys/net/ipv4/route/max_size:1524288 /proc/sys/net/ipv4/route/min_adv_mss:256 /proc/sys/net/ipv4/route/min_pmtu:552 /proc/sys/net/ipv4/route/mtu_expires:600 /proc/sys/net/ipv4/route/redirect_load:5 /proc/sys/net/ipv4/route/redirect_number:9 /proc/sys/net/ipv4/route/redirect_silence:5120 /proc/sys/net/ipv4/route/secret_interval:3600 15 sec of 30 to 50 % cpu and 15 sec 1 to 2 % cpu with /proc/sys/net/ipv4/route/gc_interval:1 almost all the time like this 01:23:45 CPU %usr %nice %sys %iowait %irq %soft %steal %guest %idle 01:23:46 all 0.00 0.00 0.00 0.00 0.00 0.12 0.00 0.00 99.88 01:23:46 0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 01:23:46 1 0.00 0.00 0.00 0.00 1.00 0.00 0.00 0.00 99.00 01:23:46 2 0.00 0.00 0.00 0.00 0.00 1.02 0.00 0.00 98.98 01:23:46 3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 01:23:46 4 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 01:23:46 5 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 01:23:46 6 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 01:23:46 7 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 with max two outputs of 20 to 30% cpu in different times from 12 to 15sec
Didn't you see any: "... rebuilds is over limit, route caching disabled" warning?
And i dont know but i think patch for turning off route cache is not working because with this patches and rt_cache_rebuild_count = 0
If you mean the patch #2, it does something opposite: with rt_cache_rebuild_count = 0 it turns off automatic "cache disabling" after rt_cache_rebuild_count events signaled with the above-mentionned warning, which was introduced in 2.6.29. Sorry for not describing this enough. Thanks, Jarek P.