feat(search): PFP-1275/look-into-0-doc-indices

[jmacryl]

reduce replicas when docs 0

[codecov]

Codecov Report

Attention: Patch coverage is 89.51049% with 15 lines in your changes missing coverage. Please review.

✅ All tests successful. No failed tests found.

Files with missing lines	Patch %	Lines
...rch/elasticsearch/indexbuilder/ESIndexBuilder.java	91.30%	3 Missing and 5 partials ⚠️
...linkedin/datahub/upgrade/system/cron/CronArgs.java	0.00%	4 Missing ⚠️
.../java/com/linkedin/datahub/upgrade/UpgradeCli.java	0.00%	2 Missing ⚠️
.../datahub/upgrade/system/cron/SystemUpdateCron.java	80.00%	1 Missing ⚠️

📢 Thoughts on this report? Let us know!

[david-leifker]

The replica count will constantly be changed to >1 on every upgrade and then downgraded by the planned cron job to adjust the replicas down. The logic to scale down to 0 also needs to be considered then a reindex is considered or it needs to ignore the diff between the intended replica count and the actual (delegating entirely to the cron).

[jmacryl]

The logic to scale down to 0 also needs to be considered then a reindex is considered or it needs to ignore the diff between the intended replica count and the actual (delegating entirely to the cron).

sorry I dont understdand this sentece, could you elaborate?

[david-leifker]

The replica count is configured at upgrade/install time as well as this new cron job. This needs to be common for both places so that the replica count doesn't get changed back & forth. One approach might be to remove the replica setting on index creation, however code needs to be updated to remove it from consideration there.

[alwaysmeticulous]

✅ Meticulous spotted visual differences in 23 of 762 screens tested, but all differences have already been approved: view differences detected.

Meticulous evaluated ~10 hours of user flows against your PR.

_{Last updated for commit 3ce4a32. This comment will update as new commits are pushed.}