Igniters, Presently every Apache Ignite node requests 80% of RAM from an operating system. We debated a lot about this default [1] and as a practice shows those who voted for the lower value (20%-25%) were right.
Our users started reporting about bizarre issues ([2], [3], [4]) when a cluster hangs, freezes for a while or even crashes during standard preloading phase or minor workloads. It’s turned out to be that lack of RAM causes extensive swapping and checkpointing activity. Even more, the issue can be reproduced on your laptop if you launch a couple of nodes and start data preloading. All this can be avoided if you allocated less than 80% and tune the checkpointing if the persistence is used. Unfortunately, only few could get to this point, the rest would kick Ignite out from their deployments. Thus, we have to fix the defaults ([5] and [6]) on our own and release an emergency release with the fixes as soon as possible. Any objections or other thoughts? [1] http://apache-ignite-developers.2346864.n4.nabble.com/IGNITE-5717-improvements-of-MemoryPolicy-default-size-td20264.html [2] http://apache-ignite-users.70518.x6.nabble.com/Performance-of-persistent-store-too-low-when-bulb-loading-tp16247p16337.html [3] http://apache-ignite-users.70518.x6.nabble.com/Strange-problems-with-Ignite-native-Persistence-when-Data-exceeds-Memory-td16187.html#a16194 [4] http://apache-ignite-users.70518.x6.nabble.com/Activation-slow-and-Ignite-node-crashed-in-the-middle-of-checkpoint-td16144.html [5] https://issues.apache.org/jira/browse/IGNITE-6003 [6] https://issues.apache.org/jira/browse/IGNITE-6182 — Denis
