[
https://issues.apache.org/jira/browse/HDDS-1561?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Nanda kumar updated HDDS-1561:
------------------------------
Description:
Right now, if a pipeline is destroyed by SCM, all the container on the pipeline
are marked as quasi closed when datanode received close container command. SCM
while processing these containers reports, marks these containers as closed
once majority of the nodes are available.
This is however not a sufficient condition in cases where the raft log
directory is missing or corrupted. As the containers will not have all the
applied transaction.
To solve this problem, we should QUASI_CLOSE the containers in datanode as part
of ratis groupRemove. If a container is in OPEN state in datanode without any
active pipeline, it will be marked as Unhealthy while processing close
container command.
cc [~jnp], [~shashikant], [~sdeka], [~nandakumar131]
was:
Right now, if a pipeline is destroyed by SCM, all the container on the pipeline
are marked as quasi closed. SCM while processing these containers reports,
marks these containers as closed once majority of the nodes are available.
This is however not a sufficient condition in cases where the raft log
directory is missing or corrupted. As the containers will not have all the
applied transaction. To solve this problem,
a new container replica state needs to be added to differentiate this from
quasi closed containers.
cc [~jnp], [~shashikant], [~sdeka], [~nandakumar131]
> Mark OPEN containers as QUASI_CLOSED as part of Ratis groupRemove
> -----------------------------------------------------------------
>
> Key: HDDS-1561
> URL: https://issues.apache.org/jira/browse/HDDS-1561
> Project: Hadoop Distributed Data Store
> Issue Type: Bug
> Components: Ozone Datanode, SCM
> Affects Versions: 0.3.0
> Reporter: Mukul Kumar Singh
> Assignee: Nanda kumar
> Priority: Blocker
>
> Right now, if a pipeline is destroyed by SCM, all the container on the
> pipeline are marked as quasi closed when datanode received close container
> command. SCM while processing these containers reports, marks these
> containers as closed once majority of the nodes are available.
> This is however not a sufficient condition in cases where the raft log
> directory is missing or corrupted. As the containers will not have all the
> applied transaction.
> To solve this problem, we should QUASI_CLOSE the containers in datanode as
> part of ratis groupRemove. If a container is in OPEN state in datanode
> without any active pipeline, it will be marked as Unhealthy while processing
> close container command.
> cc [~jnp], [~shashikant], [~sdeka], [~nandakumar131]
--
This message was sent by Atlassian JIRA
(v7.6.14#76016)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
