Hi,

I think this is a SIPP quirk. The one ‘call’ is the run of SIPP - the scenario 
of REGISTERs and INVITEs in the call_load2.xml script. As the test is 
terminated, this one call never finishes – so both successful and failed calls 
are 0.

How many subscribers were you using in your stress test?

Ellie

From: Doctor Mescaline [mailto:[email protected]]
Sent: 13 January 2015 08:53
To: Eleanor Merry
Cc: [email protected]
Subject: Re: [Clearwater] Problem with Stress testing

Hi,

if you look at the statistics under the log folder, you can see that there is 
no succesful calls and no failed calls, while there is the total number of 
calls that is greater than zero. Do you think it’s just a problem in the 
computation of the statistics?

thanks

Il giorno 12/gen/2015, alle ore 18:48, Eleanor Merry 
<[email protected]<mailto:[email protected]>> ha scritto:

Hi,

Thanks for the sending the logs over. They look OK – I’m not seeing any 
timeouts/unexpected errors (in the sipp.out file, you can see the table of SIP 
messages sent during an instance of the stress test, and see how many messages 
ended up being retransmitted/timed out, and whether any of the messages were 
unexpected).

I would guess therefore that your system is running correctly, and the timeout 
errors you were seeing before were due to running a greater load - it’s 
expected that some messages will time out when the system is heavily 
loaded/overloaded.

You’ll also see time outs when you start running at a high load, as our load 
monitor has a slow start.

As background, the load monitor will admit a request if there are available 
tokens in its token bucket (at the cost of one token). It also replenishes the 
number of tokens based on what the token rate is, and how long it's been since 
the token bucket was last replenished. Every 20 requests, it gets the latency 
of the requests (as a smoothed mean), and compares this to the target latency 
(100ms). If it is less, then the token rate is increased. How much it's 
increased by depends on how far below the average latency is to the target 
latency.

This means that if you're going from a cold start (as you do with the stress 
tests), then there is a slow start where the load monitor ramps up the token 
replacement rate (as there is a limit as to how fast the token rate can rise) 
and so a higher rate of time outs to start with.

Ellie

From: Doctor Mescaline [mailto:[email protected]]
Sent: 12 January 2015 14:32
To: Eleanor Merry
Cc: 
[email protected]<mailto:[email protected]>
Subject: Re: [Clearwater] Problem with Stress testing

Hi,

I have attached  the logs from the stress run.

Thanks for your help





Il giorno 08/gen/2015, alle ore 20:16, Eleanor Merry 
<[email protected]<mailto:[email protected]>> ha scritto:

Hi,

The REGISTERs are reaching Bono/Sprout – this rules out connectivity problems.

I don’t see any errors in the logs you’ve shown though – they all look like 
part of a registration attempt. In a successful registration, we expect the 
client to do a REGISTER with no authentication credentials. This then gets 
rejected with a 401 (which contains a WWW-authenticate header). The client will 
then resend the REGISTER, with the appropriate authorization, and this will 
then be accepted, and a 200 OK is returned.

Can you send me more details about what errors you’re seeing when you run the 
stress tests? I’d like the logs from the stress run (in 
/var/log/clearwater-sip-stress/*), as well as the full logs from Sprout (in 
/var/log/sprout/sprout_current).

Thanks,

Ellie

_______________________________________________
Clearwater mailing list
[email protected]
http://lists.projectclearwater.org/listinfo/clearwater

Reply via email to