Support for clusters with more than 300 nodes

[aimichelle]

Is your feature request related to a problem? Please describe. Pixie best supports clusters with less than 100 nodes. Due to limitations such as pulling data across all nodes into Kelvin at once, waiting for all PEMs to respond for a query, etc, queries can be slow, timeout, or crash.

Describe the solution you'd like We would like Pixie to work smoothly on clusters with more than 300 nodes.

Some alternatives to this can include limiting PEMs to specific nodes where collecting data is most useful.

[sabideep1]

hi @aimichelle @htroisi is there any update on this feature request. And is there anything in the pipeline to make it work beyond 300 Nodes. We have some clusters running with 1000 Nodes

[etep]

Hi @sabideep1 ,

This feature request is on our radar, but we have not allocated our time to it yet. Here are two topics to investigate should you be interested:

• multiple kelvin instances to parallelize query aggregation.
• limiting the amount of data returned by a query.
• multiple metadata service instances (or optimized metadata service).

Please feel free to ping us or reach out in our community slack channel.