[
https://issues.apache.org/jira/browse/IGNITE-18433?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Kirill Tkalenko updated IGNITE-18433:
-------------------------------------
Description:
Scenarios for testing: (current storages: *TxStateStorage* and
*MvPartitionStorage*)
* During a full rebalance, the leader changes, what will happen?
I expect the current rebalance to be canceled and a new one to start with empty
storages and set initial (equal to 0) *lastAppliedIndex *and *lastAppliedTerm*.
And also what index will be sent to the new leader, if during a full
rebalancing on the recipient it will be equal to *-1*.
* What happens if a leader changes during a full rebalance that has enough log
to avoid a full rebalance?
I expect the full rebalance to be canceled, clear the storages and set the
initial (equal to 0) *lastAppliedIndex *and *lastAppliedTerm*.
And also what index will be sent to the new leader, if during a full
rebalancing on the recipient it will be equal to *-1*.
* Suppose a full rebalancing for one of the storages ended successfully, and it
has updated (not equal to -1) *lastAppliedIndex *and *lastAppliedTerm*, while
for the other we did not have time to complete the full rebalancing and the
node fell.
What will happen to both storages after the node is powered on?
I expect both stores to be cleared and have initial (equal to 0)
*lastAppliedIndex *and *lastAppliedTerm*.
Each scenario can be tested in a separate ticket
was:
Scenarios for testing: (current vaults: *TxStateStorage* and
*MvPartitionStorage*)
* During a full rebalance, the leader changes, what will happen?
I expect the current rebalance to be canceled and a new one to start with empty
storages and set initial (equal to 0) *lastAppliedIndex *and *lastAppliedTerm*.
And also what index will be sent to the new leader, if during a full
rebalancing on the recipient it will be equal to *-1*.
* What happens if a leader changes during a full rebalance that has enough log
to avoid a full rebalance?
I expect the full rebalance to be canceled, clear the storages and set the
initial (equal to 0) *lastAppliedIndex *and *lastAppliedTerm*.
And also what index will be sent to the new leader, if during a full
rebalancing on the recipient it will be equal to *-1*.
* Suppose a full rebalancing for one of the storages ended successfully, and it
has updated (not equal to -1) *lastAppliedIndex *and *lastAppliedTerm*, while
for the other we did not have time to complete the full rebalancing and the
node fell.
What will happen to both storages after the node is powered on?
I expect both stores to be cleared and have initial (equal to 0)
*lastAppliedIndex *and *lastAppliedTerm*.
Each scenario can be tested in a separate ticket
> Full rebalance testing extension
> --------------------------------
>
> Key: IGNITE-18433
> URL: https://issues.apache.org/jira/browse/IGNITE-18433
> Project: Ignite
> Issue Type: Task
> Reporter: Kirill Tkalenko
> Priority: Major
> Labels: ignite-3
> Fix For: 3.0.0-beta2
>
>
> Scenarios for testing: (current storages: *TxStateStorage* and
> *MvPartitionStorage*)
> * During a full rebalance, the leader changes, what will happen?
> I expect the current rebalance to be canceled and a new one to start with
> empty storages and set initial (equal to 0) *lastAppliedIndex *and
> *lastAppliedTerm*.
> And also what index will be sent to the new leader, if during a full
> rebalancing on the recipient it will be equal to *-1*.
> * What happens if a leader changes during a full rebalance that has enough
> log to avoid a full rebalance?
> I expect the full rebalance to be canceled, clear the storages and set the
> initial (equal to 0) *lastAppliedIndex *and *lastAppliedTerm*.
> And also what index will be sent to the new leader, if during a full
> rebalancing on the recipient it will be equal to *-1*.
> * Suppose a full rebalancing for one of the storages ended successfully, and
> it has updated (not equal to -1) *lastAppliedIndex *and *lastAppliedTerm*,
> while for the other we did not have time to complete the full rebalancing and
> the node fell.
> What will happen to both storages after the node is powered on?
> I expect both stores to be cleared and have initial (equal to 0)
> *lastAppliedIndex *and *lastAppliedTerm*.
> Each scenario can be tested in a separate ticket
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
