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.
Hello Jasen,
just a quick guess – but have you already tried with a larger value suggested from the documentation?
1 sip:10.0.0.1:5060http://10.0.0.1:5060 weight=75 1 sip:10.0.0.2:5060http://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 On Behalf Of Jasen Hall Sent: Tuesday, October 15, 2019 6:56 PM To: 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:10.0.0.1:5060http://10.0.0.1:5060 weight=3 1 sip:10.0.0.2:5060http://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.comhttp://www.invoca.com
[See the list]https://invoca.sigstr.net/uc/5cba111d3527a200469b9afd
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 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 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@lists.kamailio.org *On Behalf Of *Jasen Hall *Sent:* Tuesday, October 15, 2019 6:56 PM *To:* 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 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
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?
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.demailto: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:10.0.0.1:5060http://10.0.0.1:5060 weight=75 1 sip:10.0.0.2:5060http://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.orgmailto:sr-users-bounces@lists.kamailio.org> On Behalf Of Jasen Hall Sent: Tuesday, October 15, 2019 6:56 PM To: sr-users@lists.kamailio.orgmailto: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:10.0.0.1:5060http://10.0.0.1:5060 weight=3 1 sip:10.0.0.2:5060http://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.comhttp://www.invoca.com
[See the list]https://invoca.sigstr.net/uc/5cba111d3527a200469b9afd
-- Jasen Hall | Invoca Site Reliability Engineer
www.invoca.comhttp://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/
-
Daniel Tryba -
Henning Westerholt -
Jasen Hall