[
https://issues.apache.org/jira/browse/IGNITE-17334?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17566456#comment-17566456
]
Roman Puchkovskiy edited comment on IGNITE-17334 at 7/13/22 6:19 PM:
---------------------------------------------------------------------
It is decided to drop 'rejection' policy. We will always try to spill-out to
disk when ithe n-memory buffer overflows.
was (Author: rpuch):
It is decided to drop 'rejection' policy. We will always try to spill-out to
disk when in-memory buffer overflows
> Basic volatile RAFT log storage
> -------------------------------
>
> Key: IGNITE-17334
> URL: https://issues.apache.org/jira/browse/IGNITE-17334
> Project: Ignite
> Issue Type: Improvement
> Components: persistence
> Reporter: Roman Puchkovskiy
> Assignee: Roman Puchkovskiy
> Priority: Major
> Labels: ignite-3
> Fix For: 3.0.0-alpha6
>
>
> The storage should contain the following components:
> # An elastic (but with possible hard-limit) in-memory storage for log
> entries. On this iteration it will be on-heap data structure (for instance, a
> skip list). Later, we'll build something better (like an off-heap
> implementation, maybe using pagemem).
> # An algorithm for determining when the in-memory storage cannot accept new
> entries. On this iteration, the only supported algorithm will be simply
> limiting the store by the number of stored entries. Later we will add
> something smarter, like the total size of the stored entries and adaptive
> algorithms.
> # A policy for what to do if the store cannot accept a record. On this
> iteration, it will be just to reject the entry. Later, we can add support for
> spilling out to disk.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
