hey, is it recommended that a segment reload op be carried out for a realtime table that has around 40k segments? its distributed over 40 or so servers. the reload is needed due to a schema update

You can perform table reload or server restart

will a server restart ensure that these schema updates are propagated to older segments? i saw this message "There are xx invalid segment/s. This usually means that they were created with an older schema. Please reload the table in order to refresh these segments to the new schema" seeing a message like this. • table reload: is this okay to do for tables with a mid-large number of segments like ours (any considerations wrt zookeeper or read/write path?)

yes

More reliable setups do rolling restart instead of reload. It's hard to gauge how expensive a reload will be since there’s so many degrees of freedom

have you tried this out earlier? i have some reservations about doing a rolling restart and it impacting the control path

you can limit helix transactions if it would put too much pressure on the controller side

have you tried this out earlier?

this is how we operate all of our clusters. It’s much easier to scale up 2-3 controllers to handle any load than it is to tune local reloads on servers

cool, thanks!