[
https://issues.apache.org/jira/browse/HDDS-15870?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Devesh Kumar Singh resolved HDDS-15870.
---------------------------------------
Resolution: Duplicate
> Make WritableECContainerProvider tier-aware with policy-level fallback and
> per-tier quota
> -----------------------------------------------------------------------------------------
>
> Key: HDDS-15870
> URL: https://issues.apache.org/jira/browse/HDDS-15870
> Project: Apache Ozone
> Issue Type: Sub-task
> Reporter: Devesh Kumar Singh
> Assignee: Devesh Kumar Singh
> Priority: Major
>
> {code:java}
> WritableECContainerProvider.getContainer(size, ec, owner, exclude)
> {code}
> is the entry point for allocating an EC block. Today it is fully tier-blind:
> no StorageTier parameter, no filter in the reuse loop, and a single
> getMaximumPipelines(ec) quota shared across all tiers. On a heterogeneous
> cluster this means every HOT EC allocation can end up on a DISK pipeline even
> when SSD DNs are idle, and a busy DISK tier at the global cap can starve a
> scarce SSD tier.
> Make the provider tier-aware end-to-end. Concretely, getContainer takes a
> StorageTier tier argument and internally implements the following behavior:
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
