From: "Paweł Staszewski" <pstaszewski@itcare.pl>
To: Jarek Poplawski <jarkao2@gmail.com>
Cc: Eric Dumazet <dada1@cosmosbay.com>,
Eric Dumazet <eric.dumazet@gmail.com>,
Linux Network Development list <netdev@vger.kernel.org>
Subject: Re: weird problem
Date: Wed, 15 Jul 2009 22:15:42 +0200 [thread overview]
Message-ID: <4A5E38EE.2090405@itcare.pl> (raw)
In-Reply-To: <20090714162425.GA3090@ami.dom.local>
Jarek Poplawski pisze:
> On Tue, Jul 14, 2009 at 01:26:46AM +0200, Paweł Staszewski wrote:
>
>> Jarek Poplawski pisze:
>>
>>> On Fri, Jul 10, 2009 at 04:47:54PM +0200, Jarek Poplawski wrote:
>>>
>>>
>>>> On Fri, Jul 10, 2009 at 01:59:00AM +0200, Paweł Staszewski wrote:
>>>>
>>>>
>>>>> 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?
>
With the same settings and 2.6.28 there was always cpu load from 1% to 3%
with gc_timeout = 15
>
>> 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?
>
>
Yes.
>> 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?
>
>
No i don't any info.
>> 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.
>
>
>
So is there some patch or there will be patch that turn off definitely
route cache ?
For now i use
gc_timeout = 1 in my routers and all is working fine - there is only 1
second of 20% of cpu load after every 20 sec.
Regards
Pawel Staszewski
next prev parent reply other threads:[~2009-07-15 20:15 UTC|newest]
Thread overview: 28+ messages / expand[flat|nested] mbox.gz Atom feed top
2009-06-25 16:06 weird problem Paweł Staszewski
2009-06-25 16:33 ` Paweł Staszewski
2009-06-25 17:18 ` Paweł Staszewski
2009-06-25 19:45 ` Paweł Staszewski
2009-06-25 20:18 ` Eric Dumazet
2009-06-25 22:23 ` Paweł Staszewski
2009-06-26 8:37 ` Jarek Poplawski
2009-06-26 9:05 ` Jarek Poplawski
2009-06-26 10:19 ` Eric Dumazet
2009-06-26 17:45 ` Paweł Staszewski
2009-06-26 17:57 ` Paweł Staszewski
2009-06-30 6:40 ` Jarek Poplawski
2009-06-30 8:35 ` Paweł Staszewski
2009-06-30 8:36 ` Paweł Staszewski
2009-07-08 22:34 ` Jarek Poplawski
2009-07-09 23:14 ` Paweł Staszewski
2009-07-09 23:59 ` Paweł Staszewski
2009-07-10 14:47 ` Jarek Poplawski
2009-07-11 6:24 ` Jarek Poplawski
2009-07-13 23:26 ` Paweł Staszewski
2009-07-14 16:24 ` Jarek Poplawski
2009-07-15 20:15 ` Paweł Staszewski [this message]
2009-07-15 22:43 ` Jarek Poplawski
2009-07-16 11:01 ` Jarek Poplawski
-- strict thread matches above, loose matches on Subject: below --
2003-10-14 11:00 Weird problem Jean-Rene Cormier
[not found] ` <3F8BEAEB.1060005@Loudoun-Fairfax.com>
[not found] ` <1066136413.12935.43.camel@forbidden.cipanb.ca>
2003-10-14 15:31 ` Jeffrey Laramie
[not found] ` <3F8C1700.3070902@Loudoun-Fairfax.com>
2003-10-14 16:59 ` Jean-Rene Cormier
2003-10-14 17:49 ` Jeffrey Laramie
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=4A5E38EE.2090405@itcare.pl \
--to=pstaszewski@itcare.pl \
--cc=dada1@cosmosbay.com \
--cc=eric.dumazet@gmail.com \
--cc=jarkao2@gmail.com \
--cc=netdev@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.