[
https://issues.apache.org/jira/browse/IGNITE-18603?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Kirill Tkalenko updated IGNITE-18603:
-------------------------------------
Description:
If any partition storage (*MvPartitionStorage* or *TxStateStorage*) did not
have time to rebalance, then at the start of the group raft for this partition
we must clean all its storage facilities. In general, this problem can be
present for persistent storages.
For now, the following idea seems to work:
* start storages as is;
* if one of the storages has not finished rebalancing
(*MvPartitionStorage#lastAppliedIndex ==
MvPartitionStorage#REBALANCE_IN_PROGRESS* {color:red}OR{color}
*TxStateStorage#lastAppliedIndex == TxStateStorage#REBALANCE_IN_PROGRESS*),
then we clear all the storages.
I propose in this ticket to prepare storages in another to integrate with
*org.apache.ignite.internal.table.distributed.TableManager*.
was:
If any partition storage (*MvPartitionStorage* or *TxStateStorage*) did not
have time to rebalance, then at the start of the group raft for this partition
we must clean all its storage facilities. In general, this problem can be
present for persistent storages.
For now, the following idea seems to work:
* start storages as is;
* if one of the storages has not finished rebalancing
(*MvPartitionStorage#lastAppliedIndex ==
MvPartitionStorage#REBALANCE_IN_PROGRESS* {color:red}OR{color}
*TxStateStorage#lastAppliedIndex == TxStateStorage#REBALANCE_IN_PROGRESS*),
then we clear all the storages.
> Clear all storages of a partition if one of the storages did not have time to
> rebalance
> ---------------------------------------------------------------------------------------
>
> Key: IGNITE-18603
> URL: https://issues.apache.org/jira/browse/IGNITE-18603
> Project: Ignite
> Issue Type: Task
> Reporter: Kirill Tkalenko
> Assignee: Kirill Tkalenko
> Priority: Major
> Labels: ignite-3
> Fix For: 3.0.0-beta2
>
>
> If any partition storage (*MvPartitionStorage* or *TxStateStorage*) did not
> have time to rebalance, then at the start of the group raft for this
> partition we must clean all its storage facilities. In general, this problem
> can be present for persistent storages.
> For now, the following idea seems to work:
> * start storages as is;
> * if one of the storages has not finished rebalancing
> (*MvPartitionStorage#lastAppliedIndex ==
> MvPartitionStorage#REBALANCE_IN_PROGRESS* {color:red}OR{color}
> *TxStateStorage#lastAppliedIndex == TxStateStorage#REBALANCE_IN_PROGRESS*),
> then we clear all the storages.
> I propose in this ticket to prepare storages in another to integrate with
> *org.apache.ignite.internal.table.distributed.TableManager*.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
