[
https://issues.apache.org/jira/browse/IGNITE-22876?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Roman Puchkovskiy updated IGNITE-22876:
---------------------------------------
Description:
Either a specific manager (like CmgDisasterRecoveryManager) or a method in
ClusterManagementGroupManager needs to be added.
The method signature should look like
CompletableFuture<Void> resetCluster(List<String> newCmgConsistentIds)
# It accepts consistent IDs of the members forming the new CMG
# It builds a ResetClusterMessage containing newCmgConsistentIds, existing
cluster name and a newly generated random cluster ID
# It then sends this message to each node in the physical topology (including
itself)
# Receiving this message, a node saves its contents to the Vault, responds
with OK and initiates a restart of itself
# The node on which the method is called (the Conductor) waits for all nodes
to either respond with OK or leave failing to receive the message
# If the wait finishes successfully without triggering a timeout, the
Conductor restarts itself
# If a timeout passes, but the wait above has not finished, then:
## If the majority of the new CMG nodes has replied with OK, then the
Conductor restarts itself
## Otherwise, the result is indeterminate, an exception is returned, and the
user has to manually restart the nodes that did not respond.
was:
Either a specific manager (like CmgDisasterRecoveryManager) or a method in
ClusterManagementGroupManager needs to be added.
The method signature should look like
CompletableFuture<Void> resetCluster(List<String> newCmgConsistentIds)
# It accepts consistent IDs of the members forming the new CMG
# It builds a ResetClusterMessage containing newCmgConsistentIds, existing
cluster name and a newly generated random cluster ID
# It then sends this message to each node in the physical topology (including
itself)
# Receiving this message, a node saves its contents to the Vault, responds
with OK and initiates a restart of itself
# The node on which the method is called (the Conductor) waits for either all
nodes respond with OK or fail the message send future due to leaving the
topology, or for a timeout to expire.
# If after that it turns out that the majority
> Initiate cluster reset
> ----------------------
>
> Key: IGNITE-22876
> URL: https://issues.apache.org/jira/browse/IGNITE-22876
> Project: Ignite
> Issue Type: Improvement
> Reporter: Roman Puchkovskiy
> Priority: Major
> Labels: ignite
>
> Either a specific manager (like CmgDisasterRecoveryManager) or a method in
> ClusterManagementGroupManager needs to be added.
> The method signature should look like
> CompletableFuture<Void> resetCluster(List<String> newCmgConsistentIds)
> # It accepts consistent IDs of the members forming the new CMG
> # It builds a ResetClusterMessage containing newCmgConsistentIds, existing
> cluster name and a newly generated random cluster ID
> # It then sends this message to each node in the physical topology
> (including itself)
> # Receiving this message, a node saves its contents to the Vault, responds
> with OK and initiates a restart of itself
> # The node on which the method is called (the Conductor) waits for all nodes
> to either respond with OK or leave failing to receive the message
> # If the wait finishes successfully without triggering a timeout, the
> Conductor restarts itself
> # If a timeout passes, but the wait above has not finished, then:
> ## If the majority of the new CMG nodes has replied with OK, then the
> Conductor restarts itself
> ## Otherwise, the result is indeterminate, an exception is returned, and the
> user has to manually restart the nodes that did not respond.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
