07-28-2015 01:45 PM - edited 03-18-2019 11:35 AM
Hi guys. I have calls running between 2 voip systems trough sip and i have about 7 hop between them (2 mpls and the rest of them internal between routers (fiber connection) ) .
When I see logs I see too big jitter from 1000-4000 ms , packet loss 15%-44% . I think I need to check each of the hop to find where I lose packets or earn jitter . How can I check it with standard (ping, tracert, show int status ... etc) /non-standard methods ?
I'd appreciate for any advice .
PS. routers 3750, wan 3925 .
07-28-2015 04:04 PM
Have you got end to end qos enabled? Check your QoS on edge routers for any drops etc.
Ping, traceroute, network monitoring tools, checking logs etc on all devices, packet drops etc. are some tools you can use. Traceroute end to end should give you a fair idea of which hop has latency. You can troubleshoot that segment. If you have got any network monitoring tools, you can use that as well.
-Terry
Please rate all helpful posts
07-28-2015 05:25 PM
Probably one of the easiest things to verify when it comes to e2e QoS, is setting up a packet capture using Wireshark and verify if your incoming DSCP values are ef . At least that proves that your markings are honoured all the way.
07-28-2015 04:24 PM
Just adding to what my friend Terry has advised.
Here is what I would want to do:
1. Make sure with the WAN team team there is end to end QoS Enabled (i.e. from the switchport of the PC to the Switchport of the Server it should be enabled and same set of priorities should be given to similar traffic.
2. If the QoS has had been already enabled, you want to make sure where the packet drops are happening (I believe it should be at the WAN Interface of the Router from your LAN side)? show interface should tell you if there has been any congestion and packet drops are happening
3. If there is no end to end QoS has been enabled , then there needs to be a strategy of the end to end QoS being enabled.
Cisco SRND should give detailed information about how to proceed about this.
Please let us know if there is anything more we can do for you
Regards
07-28-2015 10:20 PM
Hi,
I think the guys shared a great information about what needs to be done, especially the QoS portion. Lets look at it from troubleshooting perspective assuming that all QoS configs are in place.
1. The first tool I would use is IP SLA Echo operation between edge devices. This will provide a great statistics about each hop in the path. Based on the result, you will see at which hop the stats will start deteriorating. This will be the hop which is causing the problem (at least).
2. On the culprit hop, the first thing you need to look at is the QoS policy usage using 'show policy-map interface x/x'.
3. If the QoS looked good, the next step is CPU usage using 'show proc cpu sort'.
4. The next point to look at is speed/duplex settings to make sure that you don't have mismatch which is causing drops.
These are the main elements which contributes to such a problem. Once you fix it on the culprit node, you need to run the Echo Path operation again to see if you need to check another node.
07-29-2015 12:16 PM
Thank you for reply , i tested with ip sla and see no lost packets and ping response btw edges in 12 ms only. the big problem is QoS setting are messed up, some places has auto qos , other with class-map and policy map. I need to dig in to it with no impact to service :) .
there is auto qos on 3750 BUT no auto qos/mls qos on voice vlan and SVI of that vlan , i as i found i cant setup auto-qos on port-channel/ SVI/tunnel/subinterfaces .
But how can I join that vlan, portchannel int to qos ?
07-29-2015 12:48 PM
My advice here is to remove auto qos and configure qos based on your needs. Auto qos can mess things as it adds staff which you dont need and it includes data as well, shapping, policing, etc.
07-30-2015 12:47 PM
Doesn't change of the policy-map or Auto-qos setting affect to service (telephony, data )so I can do it during the day ? Or should I schedule this at night ?
07-31-2015 02:59 AM
Hi,
I always recommend to do activities during off-hours to avoid any popups. However, if it is pure qos modifications, it won't cause any outage. It will rather cause performance degradation.
08-02-2015 10:38 PM
Hmm ... in my opinion auto qos works generally well. Catalyst QoS shouldn't be of a big issue, as long as you have setup proper trust boundaries. That means trust the QoS for the phones/servers and then trust these over the trunk links, passing unchanged to the edge routers.
Edge would be the accurate place to check whether your traffic is correctly classified, marking are honored over the MPLS, receive preferential treatment over the MPLS and are passed on without any change end-to-end. Check for drops etc. between your edge routers. You can also used NBAR or protocol based matching on edge devices to make sure you are catching all voice traffic and correctly classifying/prioritizing. So would suggest test and fine tune qos on WAN edge routers.
-Terry
Please rate all helpful posts.
07-30-2015 03:38 PM
For IP SLA you will need to set the tos bits appropriately to math the dscp for audio and signalling to get the accurate results.
-Terry
Please rate all helpful posts
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide