[
https://issues.apache.org/jira/browse/HDDS-15760?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Siyao Meng updated HDDS-15760:
------------------------------
Release Note: The Ozone Manager no longer fails to start when a snapshot
local data YAML file is corrupt or has a snapshotId that doesn't match its
path. Such files are now logged and skipped during startup; the affected
snapshot is dropped from the in-memory version graph and OM continues to start.
The SnakeYAML parse limit for these files was also raised to 128 MiB so large
valid files load successfully. (was: Previously, if any snapshot local data
YAML file was corrupt/unparseable, or its stored snapshotId did not match its
filename/path, the Ozone Manager failed to start (init threw an IOException),
blocking startup until the offending file was manually removed. The OM now logs
and skips such files during startup instead of aborting. Additionally, the
SnakeYAML parser code-point limit for snapshot local data was raised to 128 MiB
(from the 3 MiB default) so that large, valid snapshot local data files with
many SST entries load successfully.)
> Handle large or invalid OM snapshot local data YAMLs
> ----------------------------------------------------
>
> Key: HDDS-15760
> URL: https://issues.apache.org/jira/browse/HDDS-15760
> Project: Apache Ozone
> Issue Type: Improvement
> Reporter: Siyao Meng
> Assignee: Siyao Meng
> Priority: Major
>
> OM snapshot local data YAML stores SST metadata used by snapshot defrag and
> snapshot diff. For snapshots with many small SST files, this YAML can grow
> beyond SnakeYAML's default 3 MiB code-point limit. When that happens, loading
> the YAML can fail during OM startup or snapshotLocalData initialization.
> Goal: A bad local YAML should not make OM startup fail.
> The YAML is local optimization metadata for snapshot SST/version tracking,
> not the authoritative OM namespace state. If one snapshot's local metadata is
> unreadable or inconsistent, OM should continue loading valid snapshot
> metadata and leave the affected snapshot's defrag/diff path unavailable until
> the YAML is fixed.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
