Review perf impact of gp rebuild on workspace caching, backup performance, etc.

[loujaybee]

Currently gp rebuild encourages users to update their images. However, we need to review the potential impact this has on workspace image caching, particularly as the images users use become more unique, the cache hits should decrease. We need to understand the impact this has on workspace performance. This issue is to gather data on the impact, so that we can make any decisions (or not) on how to modify, or improve either workspace caching, or the gp rebuild feature.

[loujaybee]

@gitpod-io/engineering-workspace -> Any tips and pointers to this data, or support in gathering and analysing the data would be useful.

[loujaybee]

Adding to @gitpod-io/engineering-workspace inbox for visibility, and input, but the IDE team will continue to own and drive the investigation and data analysis work for now.

[kylos101]

encourages users to update their images

How is gp rebuild doing? Can you share an example?

However, we need to review the potential impact this has on workspace image caching, particularly as the images users use become more unique, the cache hits should decrease.

I assume you are referring to workspace node and IPFS cache misses, and not a cache miss in-workspace using Docker in Workspace?

We need to understand the impact this has on workspace performance.

Less cache hits will increase start times. We have several SLOs here, for example.

the IDE team will continue to own and drive the investigation and data analysis work for now.

Given ☝️ adding to IDE team inbox.

edit: @laushinka for 👀

[loujaybee]

Thanks, @kylos101 ! 🙏

How is gp rebuild doing? Can you share an example?

We're are not doing this (nudging users to update workspace images) yet. However, a natural consequence of the feature might be more custom images, as it makes the configuration loop simpler. In future we (after a clean up of workspace images) we might get more pushy nudging users to different workspace images.

I assume you are referring to workspace node and IPFS cache misses, and not a cache miss in-workspace using Docker in Workspace?

Yep, general workspace cache is the question, not in workspace cache. However in workspace cache I presume will affect backup sizes, which is another dimension/topic of the issue 🙏

[loujaybee]

Dropping, will pick up in future.

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.