The current Search Update pipeline [FIXME:link to diagram] uses about 1/2 of the capacity of the JobRunners . When we deploy the new Search Update Pipeline, we'll be hitting the mw-api-int-async-ro API instead. Per T342252 , cirrus-streaming updater will access the mwapi through mw-on-k8s. As mentioned in the linked ticket, we need to estimate the number of rps the cirrus streaming updater will use. (See this dashboard for an example of a similar flink app causing 20x RPS increase during its backfill).
Creating this ticket to:
Estimate the number of RPS the streaming updater will use.Determine if there are other important usage metrics (payload size?) and estimate these as well.