15 Oct
2019
15 Oct
'19
6:55 p.m.
I am trying to change our request dispatching from round-robin to a
weight-based method (ds_select_dst() algorithm 9 vs. algorithm 4) and am
having a hard time getting an asymmetrical dispatch pattern enabled. In
testing, I'm dispatching to two Ringswitch servers and trying to send three
times the traffic to one instance.
My destinations list file previously set both endpoints in the same group
and simply provided the SIP URI for each. My new list file adds the
attribute `weight` to each endpoint and looks like this:
1 sip:10.0.0.1:5060 weight=3
1 sip:10.0.0.2:5060 weight=1
However, when I throw a couple thousand calls at Kamailio, I always end up
with an almost perfect split of calls to both ringswitches. Regardless of
the weights I choose, I never see variation in the dispatched load.
Does anyone have experience with this issue and have some guidance?
Thanks in advance.
--
Jasen Hall | Invoca
*Site Reliability Engineer*
www.invoca.com
[image: See the list]
<https://invoca.sigstr.net/uc/5cba111d3527a200469b9afd>
16 Oct
16 Oct
12:15 p.m.
Hello Jasen,
just a quick guess – but have you already tried with a larger value suggested from the
documentation?
1 sip:10.0.0.1:5060<http://10.0.0.1:5060>
weight=75
1 sip:10.0.0.2:5060<http://10.0.0.2:5060> weight=25
Cheers,
Henning
--
Henning Westerholt - https://skalatan.de/blog/
Kamailio services - https://skalatan.de/services
From: sr-users <sr-users-bounces(a)lists.kamailio.org> On Behalf Of Jasen Hall
Sent: Tuesday, October 15, 2019 6:56 PM
To: sr-users(a)lists.kamailio.org
Subject: [SR-Users] Kamailio weighted dispatch problems
I am trying to change our request dispatching from round-robin to a weight-based method
(ds_select_dst() algorithm 9 vs. algorithm 4) and am having a hard time getting an
asymmetrical dispatch pattern enabled. In testing, I'm dispatching to two Ringswitch
servers and trying to send three times the traffic to one instance.
My destinations list file previously set both endpoints in the same group and simply
provided the SIP URI for each. My new list file adds the attribute `weight` to each
endpoint and looks like this:
1 sip:10.0.0.1:5060<http://10.0.0.1:5060>
weight=3
1 sip:10.0.0.2:5060<http://10.0.0.2:5060> weight=1
However, when I throw a couple thousand calls at Kamailio, I always end up with an almost
perfect split of calls to both ringswitches. Regardless of the weights I choose, I never
see variation in the dispatched load.
Does anyone have experience with this issue and have some guidance?
Thanks in advance.
--
Jasen Hall | Invoca
Site Reliability Engineer
www.invoca.com<http://www.invoca.com>
[See the list]<https://invoca.sigstr.net/uc/5cba111d3527a200469b9afd>
5:51 p.m.
Thanks for the reply, Henning.
I have tried an embarrassing variety of weights thinking I was
misinterpreting the instructions. I've tried weight divisions of: 1/3,
10/30, 90/9, 90/10,100/0, 90/30, and more; all with the hopes of generating
any variance in a straight split of the traffic. In every case I get an
almost equal split of calls dispatched.
On Wed, Oct 16, 2019 at 3:15 AM Henning Westerholt <hw(a)skalatan.de> wrote:
Hello Jasen,
just a quick guess – but have you already tried with a larger value
suggested from the documentation?
--
Jasen Hall | Invoca
*Site Reliability Engineer*
www.invoca.com
[image: See the list]
<https://invoca.sigstr.net/uc/5cba111d3527a200469b9afd>
1 sip:10.0.0.1:5060 weight=75
1 sip:10.0.0.2:5060 weight=25
Cheers,
Henning
--
Henning Westerholt - https://skalatan.de/blog/
Kamailio services - https://skalatan.de/services
*From:* sr-users <sr-users-bounces(a)lists.kamailio.org> *On Behalf Of *Jasen
Hall
*Sent:* Tuesday, October 15, 2019 6:56 PM
*To:* sr-users(a)lists.kamailio.org
*Subject:* [SR-Users] Kamailio weighted dispatch problems
I am trying to change our request dispatching from round-robin to a
weight-based method (ds_select_dst() algorithm 9 vs. algorithm 4) and am
having a hard time getting an asymmetrical dispatch pattern enabled. In
testing, I'm dispatching to two Ringswitch servers and trying to send three
times the traffic to one instance.
My destinations list file previously set both endpoints in the same group
and simply provided the SIP URI for each. My new list file adds the
attribute `weight` to each endpoint and looks like this:
1 sip:10.0.0.1:5060 weight=3
1 sip:10.0.0.2:5060 weight=1
However, when I throw a couple thousand calls at Kamailio, I always end up
with an almost perfect split of calls to both ringswitches. Regardless of
the weights I choose, I never see variation in the dispatched load.
Does anyone have experience with this issue and have some guidance?
Thanks in advance.
--
Jasen Hall | Invoca
*Site Reliability Engineer*
www.invoca.com
[image: See the list]
<https://invoca.sigstr.net/uc/5cba111d3527a200469b9afd>
6:08 p.m.
On Wed, Oct 16, 2019 at 08:51:35AM -0700, Jasen Hall wrote:
I have tried an embarrassing variety of weights
thinking I was
misinterpreting the instructions. I've tried weight divisions of: 1/3,
10/30, 90/9, 90/10,100/0, 90/30, and more; all with the hopes of generating
any variance in a straight split of the traffic. In every case I get an
almost equal split of calls dispatched.
Well, maybe the problem isn't in the weights but your calls to
ds_select_*?
Using rweight with alog 11 I can say it works. So double check the
ds_select calls?
17 Oct
17 Oct
8:59 a.m.
Hi Jasen,
I see. if you can confirm that you call it correctly in the ds_select function call - then
please open a bug report about it in our github issue tracker.
Cheers,
Henning
Am 16.10.19 um 17:51 schrieb Jasen Hall:
Thanks for the reply, Henning.
I have tried an embarrassing variety of weights thinking I was misinterpreting the
instructions. I've tried weight divisions of: 1/3, 10/30, 90/9, 90/10,100/0, 90/30,
and more; all with the hopes of generating any variance in a straight split of the
traffic. In every case I get an almost equal split of calls dispatched.
On Wed, Oct 16, 2019 at 3:15 AM Henning Westerholt
<hw@skalatan.de<mailto:hw@skalatan.de>> wrote:
Hello Jasen,
just a quick guess – but have you already tried with a larger value suggested from the
documentation?
1 sip:10.0.0.1:5060<http://10.0.0.1:5060>
weight=75
1 sip:10.0.0.2:5060<http://10.0.0.2:5060> weight=25
Cheers,
Henning
--
Henning Westerholt - https://skalatan.de/blog/
Kamailio services - https://skalatan.de/services
From: sr-users
<sr-users-bounces@lists.kamailio.org<mailto:sr-users-bounces@lists.kamailio.org>>
On Behalf Of Jasen Hall
Sent: Tuesday, October 15, 2019 6:56 PM
To: sr-users@lists.kamailio.org<mailto:sr-users@lists.kamailio.org>
Subject: [SR-Users] Kamailio weighted dispatch problems
I am trying to change our request dispatching from round-robin to a weight-based method
(ds_select_dst() algorithm 9 vs. algorithm 4) and am having a hard time getting an
asymmetrical dispatch pattern enabled. In testing, I'm dispatching to two Ringswitch
servers and trying to send three times the traffic to one instance.
My destinations list file previously set both endpoints in the same group and simply
provided the SIP URI for each. My new list file adds the attribute `weight` to each
endpoint and looks like this:
1 sip:10.0.0.1:5060<http://10.0.0.1:5060>
weight=3
1 sip:10.0.0.2:5060<http://10.0.0.2:5060> weight=1
However, when I throw a couple thousand calls at Kamailio, I always end up with an almost
perfect split of calls to both ringswitches. Regardless of the weights I choose, I never
see variation in the dispatched load.
Does anyone have experience with this issue and have some guidance?
Thanks in advance.
--
Jasen Hall | Invoca
Site Reliability Engineer
www.invoca.com<http://www.invoca.com>
[See the
list]<https://invoca.sigstr.net/uc/5cba111d3527a200469b9afd>
--
Jasen Hall | Invoca
Site Reliability Engineer
www.invoca.com<http://www.invoca.com>
[See the list]<https://invoca.sigstr.net/uc/5cba111d3527a200469b9afd>
--
Kamailio Merchandising - https://skalatan.de/merchandising/
Kamailio services - https://skalatan.de/services
Henning Westerholt - https://skalatan.de/blog/
1676
days inactive
1678
days old
4 comments
3 participants
participants (3)
-
Daniel Tryba -
Henning Westerholt -
Jasen Hall