Netflix / inviso

Other
205 stars 65 forks source link

too many data points #18

Open synfinatic opened 7 years ago

synfinatic commented 7 years ago

Getting the following error even for relatively small (sub day long) selections:

2017-02-10 08 16 10 am

danielcweeks commented 7 years ago

The limit is here: https://github.com/Netflix/inviso/blob/master/web-ui/public/js/cluster.js#L192

The purpose of the limit is to ensure that the results can be retrieved in a reasonable amount of time and D3 can render it. My guess is that you have lots of small applications, which means there would be too many data points to render in the application view.

There are a couple ways that you can handle this. One would be to just increase the limit above. Another would be to reduce your indexing frequency. If you call index_cluster_stats.py too frequently, you might have too many values. Make sure that you only index every minute and if it's too small drop it down to every two minutes.

In our clusters we can typically only look at about 2 hours worth of data with application level detail, but it all depends on how many applications were running during that time.

Let me know if that helps.