[HDDS-6234] Repair containers affected by incorrect used bytes and block count

• Type: New Feature
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.2.0
• Fix Version/s: 1.3.0
• Component/s: Ozone Datanode, SCM
• Labels: pull-request-available
• Assignee: Ethan Rose
• Reporter: Ethan Rose
• Created: 27/Jan/22 22:34
• Updated: 27/Jun/23 17:12
• Resolved: 17/Feb/22 17:46

Description

HDDS-5359 shows that the bytes used and block count metadata of a container is tracked incorrectly. Empty containers with negative values in these fields will not be deleted per HDDS-4023 and will remain tracked by SCM unnecessarily.

This Jira aims to add a config flag that when enabled will repair the metadata information in the datanodes on restart. Datanodes already iterate all their containers on startup, but the process is kept light to not slow down startup too much. Repairing these values will slow down datanode startup, since it requires writing to RocksDB (currently opened in readonly mode on startup) and iterating the block keys stored there. For this reason the plan is to put this behind a config flag.

I have set the affected versions to 1.2.0 since it is after this release that HDDS-5359 was reported, but it is possible earlier releases are affected.

Issue Links

• depends upon: HDDS-5359 Incorrect BLOCKCOUNT and BYTESUSED in container DB
• links to: GitHub Pull Request #3042