Baseline topology without persistence

[Valentin Kulichenko]

Folks,

By default baseline topology is enforced only if persistence is enabled and
is not defined in in-memory only scenario. But does this mean that I still
CAN use it without persistence?

In some cases users want to avoid rebalancing in case of a node failure
because it's assumed that this node will come back up very quickly. For
example, there are 3 nodes and one of them dies. If data is rebalanced
right away, I would need more memory space on 2 alive nodes which I might
not have. Instead of increasing the amount of data there, I want to
resurrect the dead node and restore the missing partitions there.

Is BLT the right way to achieve behavior described above and if so, what
are the exact steps? Should I set activeOnStart to false and then operation
just like I would do with persistence?

-Val

[Stanislav Lukyanov]

Currently BLT can’t be used without persistence – only persistent caches are affected by it.
AFAIR making non-persistent caches make use of the baseline topology is the plan for IEP-4 Phase 2.

Thanks,
Stan

From: Valentin Kulichenko
Sent: 19 июля 2018 г. 0:04
To: [hidden email]
Subject: Baseline topology without persistence

Folks,

By default baseline topology is enforced only if persistence is enabled and
is not defined in in-memory only scenario. But does this mean that I still
CAN use it without persistence?

In some cases users want to avoid rebalancing in case of a node failure
because it's assumed that this node will come back up very quickly. For
example, there are 3 nodes and one of them dies. If data is rebalanced
right away, I would need more memory space on 2 alive nodes which I might
not have. Instead of increasing the amount of data there, I want to
resurrect the dead node and restore the missing partitions there.

Is BLT the right way to achieve behavior described above and if so, what
are the exact steps? Should I set activeOnStart to false and then operation
just like I would do with persistence?

-Val