You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@ignite.apache.org by "Sergey Chugunov (Jira)" <ji...@apache.org> on 2020/12/11 14:23:00 UTC
[jira] [Updated] (IGNITE-12892) Clarify WAL archive size
configuration
[ https://issues.apache.org/jira/browse/IGNITE-12892?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Sergey Chugunov updated IGNITE-12892:
-------------------------------------
Fix Version/s: 2.10
> Clarify WAL archive size configuration
> --------------------------------------
>
> Key: IGNITE-12892
> URL: https://issues.apache.org/jira/browse/IGNITE-12892
> Project: Ignite
> Issue Type: Sub-task
> Reporter: Semyon Danilov
> Assignee: Semyon Danilov
> Priority: Major
> Fix For: 2.10
>
> Time Spent: 50m
> Remaining Estimate: 0h
>
> Actual maximum size of WAL archive that can be reserved for historical rebalance is calculated as minimum of three properties:
> # DataStorageConfiguration#walHistSize (units: number of checkpoints)
> # DataStorageConfiguration#maxWalArchiveSize (units: bytes)
> # IgniteSystemProperties#IGNITE_PDS_MAX_CHECKPOINT_MEMORY_HISTORY_SIZE (units: number of checkpoints)
> The logic is a little unclear, so I propose following changes:
> # Stop using walHistSize at all (it's already deprecated) for WAL truncation
> # Use IGNITE_PDS_MAX_CHECKPOINT_MEMORY_HISTORY_SIZE only in case WAL archive is managed externally (so we limit the quantity of checkpoints stored in memory, but don't remove WAL files)
> # Use -1 for maxWalArchiveSize (instead of Long.MAX_VALUE) to disable WAL truncation
--
This message was sent by Atlassian Jira
(v8.3.4#803005)