capturing radius performance metrics
I would like to monitor the ongoing real-world performance of my freeradius server. In this case, I'm defining performance as time between receiving request and sending answer. I have used radsniff and radclient to do some benchmarking, but neither really seem like a sustainable or appropriate solution for long-term data collection in a production environment. Is there a way to add the "total response time" to the log line? or maybe there is a statistics channel like bind has to access performance metrics? I am able to collect a similar metric from the NAS as well as from some "fake" clients I configured to track radius authentication performance. I'm hoping there is a way to also gather this from the freeradius service directly. Thanks. -- Munroe Sollog Senior Network Engineer munroe@lehigh.edu
On Jan 3, 2020, at 4:15 PM, Munroe Sollog <mus3@lehigh.edu> wrote:
I would like to monitor the ongoing real-world performance of my freeradius server. In this case, I'm defining performance as time between receiving request and sending answer. I have used radsniff and radclient to do some benchmarking, but neither really seem like a sustainable or appropriate solution for long-term data collection in a production environment.
Is there a way to add the "total response time" to the log line? or maybe there is a statistics channel like bind has to access performance metrics?
I am able to collect a similar metric from the NAS as well as from some "fake" clients I configured to track radius authentication performance. I'm hoping there is a way to also gather this from the freeradius service directly.
You can user radmin to get overall statistics from the server. There's a "stats" command. Just connect via radmin and type "help". See the stats via: radmin> stats client auth .. and a number of statistics will be printed out. Alan DeKok.
On 4/01/2020, at 10:15 AM, Munroe Sollog <mus3@lehigh.edu> wrote:
I would like to monitor the ongoing real-world performance of my freeradius server. In this case, I'm defining performance as time between receiving request and sending answer. I have used radsniff and radclient to do some benchmarking, but neither really seem like a sustainable or appropriate solution for long-term data collection in a production environment.
You can have radsniff write to a collectd socket. I have this in many many production servers, it works very well. Collectd can write out many formats, including to other collection services over a network. Of course, it shows packet rate etc. stats rather than radius process internal stats, but it gets you most of the way there. -- Nathan Ward
For those of you using radsniff for stats collection, has anyone had any luck using the telgraf collectd plugin? radsniff seems to work fine to collectd but when I plumb it to telegraf, telegraf complains about invalid packet length. I have a bug open for telegraf, but I was wondering if anyone else has the same issue. https://github.com/influxdata/telegraf/tree/master/plugins/parsers/collectd On Fri, Jan 3, 2020 at 6:05 PM Nathan Ward <lists+freeradius@daork.net> wrote:
On 4/01/2020, at 10:15 AM, Munroe Sollog <mus3@lehigh.edu> wrote:
I would like to monitor the ongoing real-world performance of my freeradius server. In this case, I'm defining performance as time between receiving request and sending answer. I have used radsniff and radclient to do some benchmarking, but neither really seem like a sustainable or appropriate solution for long-term data collection in a production environment.
You can have radsniff write to a collectd socket. I have this in many many production servers, it works very well.
Collectd can write out many formats, including to other collection services over a network.
Of course, it shows packet rate etc. stats rather than radius process internal stats, but it gets you most of the way there.
-- Nathan Ward
- List info/subscribe/unsubscribe? See http://www.freeradius.org/list/users.html
-- Munroe Sollog Senior Network Engineer munroe@lehigh.edu
On 20 Jan 2020, at 09:11, Munroe Sollog <mus3@lehigh.edu> wrote:
For those of you using radsniff for stats collection, has anyone had any luck using the telgraf collectd plugin? radsniff seems to work fine to collectd but when I plumb it to telegraf, telegraf complains about invalid packet length. I have a bug open for telegraf, but I was wondering if anyone else has the same issue.
Not used it unfortunately, but how radsniff writes out the data is via the libcollectdclient library, so it's unlikely to be an issue with the code in radsniff itself. I'm glad other people are finding this functionality useful :) -Arran Arran Cudbard-Bell <a.cudbardb@freeradius.org> FreeRADIUS Development Team FD31 3077 42EC 7FCD 32FE 5EE2 56CF 27F9 30A8 CAA2
Well, it's not useful yet. I need to wrangle telegraf before I can do anything with it. On Mon, Jan 20, 2020 at 1:22 PM Arran Cudbard-Bell < a.cudbardb@freeradius.org> wrote:
On 20 Jan 2020, at 09:11, Munroe Sollog <mus3@lehigh.edu> wrote:
For those of you using radsniff for stats collection, has anyone had any luck using the telgraf collectd plugin? radsniff seems to work fine to collectd but when I plumb it to telegraf, telegraf complains about invalid packet length. I have a bug open for telegraf, but I was wondering if anyone else has the same issue.
Not used it unfortunately, but how radsniff writes out the data is via the libcollectdclient library, so it's unlikely to be an issue with the code in radsniff itself.
I'm glad other people are finding this functionality useful :)
-Arran
Arran Cudbard-Bell <a.cudbardb@freeradius.org> FreeRADIUS Development Team
FD31 3077 42EC 7FCD 32FE 5EE2 56CF 27F9 30A8 CAA2
- List info/subscribe/unsubscribe? See http://www.freeradius.org/list/users.html
-- Munroe Sollog Senior Network Engineer munroe@lehigh.edu
On 20 Jan 2020, at 14:11, Munroe Sollog <mus3@lehigh.edu> wrote:
Well, it's not useful yet.
It's ok, Nathan's message will sustain me through these cold cold Canadian winter nights. -Arran Arran Cudbard-Bell <a.cudbardb@freeradius.org> FreeRADIUS Development Team FD31 3077 42EC 7FCD 32FE 5EE2 56CF 27F9 30A8 CAA2
Use the "status" server https://wiki.freeradius.org/config/Status On Fri, Jan 3, 2020 at 1:15 PM Munroe Sollog <mus3@lehigh.edu> wrote:
I would like to monitor the ongoing real-world performance of my freeradius server. In this case, I'm defining performance as time between receiving request and sending answer. I have used radsniff and radclient to do some benchmarking, but neither really seem like a sustainable or appropriate solution for long-term data collection in a production environment.
Is there a way to add the "total response time" to the log line? or maybe there is a statistics channel like bind has to access performance metrics?
I am able to collect a similar metric from the NAS as well as from some "fake" clients I configured to track radius authentication performance. I'm hoping there is a way to also gather this from the freeradius service directly.
Thanks. -- Munroe Sollog Senior Network Engineer munroe@lehigh.edu - List info/subscribe/unsubscribe? See http://www.freeradius.org/list/users.html
On 20 Jan 2020, at 14:19, Zetan Drableg <zetan.drableg@gmail.com> wrote:
Use the "status" server https://wiki.freeradius.org/config/Status
FreeRADIUS lies, or the kernel lies or both lie. Radsniff knows the truth. Radsniff will set you free. -Arran Arran Cudbard-Bell <a.cudbardb@freeradius.org> FreeRADIUS Development Team FD31 3077 42EC 7FCD 32FE 5EE2 56CF 27F9 30A8 CAA2
participants (5)
-
Alan DeKok -
Arran Cudbard-Bell -
Munroe Sollog -
Nathan Ward -
Zetan Drableg