Sumit Agrawal created HDDS-9588:
-----------------------------------
Summary: DN import of container is not safe while replication
Key: HDDS-9588
URL: https://issues.apache.org/jira/browse/HDDS-9588
Project: Apache Ozone
Issue Type: Bug
Reporter: Sumit Agrawal
Assignee: Sumit Agrawal
Scenario:
1. SCM has triggered replicate from Src1 to Target1 DN.
2. Timeout (10 min)
3. SCM again triggered replicate from Src2 to Target1 DN (same target)
This creates possibility of parallel / sequential execution of import at target
DN.
Environment have multiple volume, as per code analysis, there are below
scenario where import can have issue at target,
Since import do not check if container is already present or any existing
import is going on for same container,
# Volume chosen for first import is V1 and import is success
# Volume chosen for second import is V2 and import will be success but add to
container set will be ignored/failed
This have issue that V2 volume will have container and consume disk space, but
never reported. This have less impact, just report unknown container.
When have same volume for both import, and operation are getting parallel for
Import at target DN,
* While un-tar the zip at volume location, there can be issue of blocks
deleted or corrupted, as operation performed,
** delete container folder from temp location
** un-tar the zip file
** move the files to destination
So in this case,
Import 1: un-tar the zip file and started move of files
Import 2: parallely deleted the container folder and un-tar in progress
This can cause missing of files and/or corrupt the files checksum.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
