RUM Configuration Tuning for High Traffic Event

[Kor03]

Hi!

We have a large marketing event coming up. We've deployed agentless RUM on the landing page, and have OneAgent injected RUM on the downstream services. I want to make sure our RUM configuration is setup to capture the burst in traffic best during the event. Lots of eyes will be on the dashboards 🙂

So far, we're making sure our CDN is ignoring the proper items in the cache rules, as well as temporarily increasing our RUM overload protection. 

Current questions I have:

1. Sampling: What's a good sampling % for a large burst? Do you account for sampling when displaying the RUM data? Example: either clearly label that the data is sampled, or multiply count values for an estimated total?
2. RUM Overload Protection: Is there a limit other than DEM consumption? 
3. Is there any further caching configuration we should consider for a large traffic burst? We want to avoid gaps in data. And we obviously do not want to introduce any performance issues on our site or on the Dynatrace side

I'll be running this all past our support team, but thought I'd ask here as well 🙂

Any other tips (or things to avoid) would be appreciated!

Thank you!

Korwin

[tyler_rowe]

1. Is the sampling to help manage DEM consumption, or for other reasons? Typically, we don't see much sampling for large campaigns/expected traffic spikes, as people typically want the full metrics. For the RUM application itself, it won't mention the sampling anywhere outside of the cost control. If you are using USQL and have a large enough volume that it starts to do extrapolation, the USQL tile will have a little ! icon on it that will mention that.

2. If you are on SaaS, there will be a soft limit to have far you can increase it, but you can work with your account team to have that increased if needed. If on Managed, you'd want to ensure the cluster (and Cluster ActiveGate for the agentless traffic) can handle the increased traffic. It isn't usually a concern unless you are expecting a significant spike.

3. Just be aware that you may get a traffic spike notification if you have traffic spike alerts on in the application. Other than that, there shouldn't be any problems on the Dynatrace side other than validating that you are not approaching your configured overload prevention limit. The only potential recommendation would be that, if you are using OneAgent on SaaS, you can turn on the Cluster ActiveGate option under the Beacon Endpoint section for your application. This will offload the beacons going back to OneAgent/origin, and just send them directly to Dynatrace, reducing traffic on your side.