2017-01-31 - Re: [GRASE-Hotspot] is coova-chilli compiled with –with-nfcoova ?

Header Data

From: Timothy White <ti***8@gmail.com>
Message Hash: f84335caa3b7cfe2e71e2464eb150e9864bfa8848405d87b20e150578f547989
Message ID: <CAESLx0+pzuV3ykFQv2vJq+sRakLGfc8C7iYE-6LvW=XkFoz_qA@mail.gmail.com>
Reply To: <a094ec0d-f289-4f94-993e-c3004938e3d7@grasehotspot.org>
UTC Datetime: 2017-01-31 02:22:13 UTC
Raw Date: Tue, 31 Jan 2017 19:22:13 +1000

Raw message 

Hi Gianluca

Can you collect support information (
https://grasehotspot.org/support/collecting-support-information/) just so I
can see the rest of the setup. It's possible that being an older Coova
Chilli version there could be a bug. What Distro are you on? I may be able
to get the newer chilli package to you in the new repo, even if other parts
aren't ready.

Regards

Tim

On Tue, Jan 31, 2017 at 2:20 AM, Gianluca Filippini <
gi***4@gmail.com> wrote:

> thanks Tim,
> I'll have to play around with the accounting interval..
>
> yeah .. I have a medium-large setup and that is the root cause of my
> problems I suppose.
> I've already tweaked the number of threads for the database and other
> variables .. but I still see this issue when I reach the peak of load.
>
> in average I have 120 users, peak of 180, generating a constant  30Mbit/s
> with a peak of 60Mbit/s of traffic.
> the machine has plenty of resources .. 16G ram, 4cores xeon 3.2Ghz .. and
> a 100Mbit/s symmetrical (FTTH)
> monitoring the hw I don;t see it to max out cpu / ram anytime
>
> still ... I'm at the point where freeradius get stuck and I need to
> restart processes (if not the whole machine) very frequently.
> when the load is lower (like 30/40 users) everything seems stable..
>
> I suspect coova/freeradius are the culprit of this problem
>
> thanks
>
>
>
> On Monday, January 30, 2017 at 12:27:18 PM UTC+1, timwhite88 wrote:
>>
>> Hi Gianluca
>>
>> Unfortunately nfcoova is a kernel module, so much harder to setup. It
>> would require that we build it targeted at a specific kernel version, on a
>> specific Linux distribution and version.
>> I am doing work on splitting the repository into multiple, one per distro
>> version, however even when that's done, it'll take lots of time for me to
>> get the nfcoova stuff working in a nice way that I can package it and have
>> it reliably run.
>>
>> Having said that, the error messages you have posted don't suggest you
>> need nfcoova, but rather you need to work out why the radius queue is full,
>> and tweak that. Monitor your server for high load, see if it's a database
>> server issue, or is it freeradius. If you look for a recent post on the
>> mailing list, you'll see another thing to try is changing the Accounting
>> Interval.
>>
>> nfcoova is only for throughput issues, so you need a large upstream pipe,
>> and enough users that you can actually utilise that pipe.
>>
>> Regards
>>
>> Tim
>>
>> On Mon, Jan 30, 2017 at 5:56 PM, Gianluca Filippini <
>> gi***.@gmail.com> wrote:
>>
>>> Hi Tim et all,
>>> one of the issues that I have on my current (old) grase installation
>>> (medium large ... ~200 users) is that quite often freeradius dies  with a
>>> message like
>>>
>>> Jan 30 08:35:06 wifi coova-chilli[1220]: radius.c: 240: radius queue is
>>>> full! qnext=83 qsize=0
>>>> Jan 30 08:35:06 wifi coova-chilli[1220]: chilli.c: 1503:
>>>> radius_default_pack() failed
>>>>
>>>
>>> so I looked for solutions and I ended up focusing on these two links:
>>>
>>> https://github.com/coova/coova-chilli/issues/61
>>> https://coova.github.io/mail-archive/chilli/2010-April/001239.html
>>>
>>> it seems that performance is an issue for large amount of traffic unless
>>> we use this module for xt-coova.
>>>
>>> is this available in the current GRASE release? it seems to me that a
>>> new compile option is needed.
>>>
>>> thanks
>>>
>>> --
>>> This mailing list is for the Grase Hotspot Project
>>> http://grasehotspot.org
>>> ---
>>> You received this message because you are subscribed to the Google
>>> Groups "Grase Hotspot" group.
>>> To unsubscribe from this group and stop receiving emails from it, send
>>> an email to gr***.@grasehotspot.org.
>>> To post to this group, send email to gr***.@grasehotspot.org.
>>> Visit this group at https://groups.google.com/a/gr
>>> asehotspot.org/group/grase-hotspot/.
>>> To view this discussion on the web visit https://groups.google.com/a/gr
>>> asehotspot.org/d/msgid/grase-hotspot/5c979d23-2f89-4af4-
>>> a7ac-a98809e7ec26%40grasehotspot.org
>>> <https://groups.google.com/a/grasehotspot.org/d/msgid/grase-hotspot/5c979d23-2f89-4af4-a7ac-a98809e7ec26%40grasehotspot.org?utm_medium=email&utm_source=footer>
>>> .
>>>
>>
>> --
> This mailing list is for the Grase Hotspot Project http://grasehotspot.org
> ---
> You received this message because you are subscribed to the Google Groups
> "Grase Hotspot" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to gr***e@grasehotspot.org.
> To post to this group, send email to gr***t@grasehotspot.org.
> Visit this group at https://groups.google.com/a/
> grasehotspot.org/group/grase-hotspot/.
> To view this discussion on the web visit https://groups.google.com/a/
> grasehotspot.org/d/msgid/grase-hotspot/a094ec0d-f289-
> 4f94-993e-c3004938e3d7%40grasehotspot.org
> <https://groups.google.com/a/grasehotspot.org/d/msgid/grase-hotspot/a094ec0d-f289-4f94-993e-c3004938e3d7%40grasehotspot.org?utm_medium=email&utm_source=footer>
> .
>

Thread