Kibana: Heatmap experiencing severe slowdown with certain data

[aaronjcaldwell]

Kibana version:

7.9

Describe the bug:

Starting in 7.9, a user that previously had no issues loading Kibana Heatmaps in 7.8 now experiences severe slowdown taking several minutes to load where it used to only take a few seconds. This issue was reported to us but I was unable to repro locally so it may be data or quantity-dependent.

I was able to inspect the chrome performance profile and confirm the issue appears to be in baseMergeDeep which appears to be used internally by lodash. Has there been a lodash merge-related change since 7.8 that might affect this?

[elasticmachine]

Pinging @elastic/kibana-app (Team:KibanaApp)

[consulthys]

I have the exact same issue with a simple multi-series bar chart that was running fine on 7.8. baseMergeDeep seems to be the culprit indeed according to Chrome performance. I used to be able to view way beyond 10 days on this histogram, but now if I select more than 2 days, it slows down beyond reason, even though the underlying query returns in milliseconds when run in Dev Tools.

[filippog]

I've been experiencing the same CPU high usage and general slowness in other types of visualizations as well with 7.9.

With a debug-enabled Kibana build I've been able to trace the deep call stacks to usages of defaultsDeep Lodash function from within the visualization library, although I don't know what changed that might have caused the regression. Hope that helps!

[wylieconlon]

Fixed by https://github.com/elastic/kibana/pull/75850, will release in 7.9.1

[consulthys]

I confirm I no longer have this issue after upgrading to 7.9.1