Implications of disabling initial state transfer?

[vdzhuvinov]

With large caches, or where JVM memory gets short, and cached objects get moved to disk swap space, occasionally we get state transfer timeouts when nodes get added or recycled.

We deal with these the usual way, by raising the state transfer timeout, sometimes more than once, until the state transfer is able to complete.

Apparently, disabling initial state transfer also works, to prevent such startup exceptions when a node gets added. I kind of like that, because it "guarantees" that node addition / recycling will always work without hitches.

But are there any hidden and potentially bad implications of disabling initial state transfer, with replicated and resp. distributed caches?

Thanks,

[dan.berindei]

Do you mean disabling await-initial-transfer? This doesn't disable the initial state transfer, it just stops getCache() from blocking to wait for that initial transfer. There should be no bad implications of doing that.

[vdzhuvinov]

Hi Dan,

I meant setting <state-transfer enable="false"/>