search

sipcapture / homer-app

HOMER 7.x Front-End and API Server
http://sipcapture.io
GNU Affero General Public License v3.0
207 stars 85 forks source link

Different Flow view for after BYE #462

Closed solarmon closed 2 years ago

solarmon commented 2 years ago

Hi,

I have a Homer 7.10.10 setup. This issues occurs on two separate Homer servers - opensips is sending HEP capture traffic to both.

The issue I'm experiencing is that, for the same callid (and correlation_id), when I click on callid of the BYE packet, it shows a different Flow view compared with when I click the callid of any packets before the BYE paket.

image

In this example, this is for a long duration call - nearly 6 hours.

And on this occasion it is just showing the flow for after the BYE.

In another occasion, when I click before the BYE it shows the Flow up to (not including) the BYE, and when I click on the BYE it shows the complete Flow including and after the BYE.

I can try updating Homer to the latest version, if that resolves the issue?

lmangani commented 2 years ago

Could you please do confirm if this is still an issue with the latest version, otherwise the patch and investigation might diverge

solarmon commented 2 years ago

I updated to Homer 7.10.15 and it seems to have the same issue for me.

solarmon commented 2 years ago

I changed the lookback/lookforward range (in Admin > Mapping > Protocol > Correlation) for "call SIP" to:

"lookup_range": [
            -6000,
            4000
        ]

And that seems to have resolved the issue.

What is the implications on performance and resources by of having a wide range?

lmangani commented 2 years ago

@solarmon the implications as you suspected is an increased resource usage per query, but you can observe the setup consumption and judge if this is a critical increase or not - being an ondemand query it should not impact you much unless dozens of users leverage it concurrently. You should find the minimal "sweet spot"

lmangani commented 2 years ago

Closing as resolved, suggestions/docs contributions are welcome on the subject!