[
https://issues.apache.org/jira/browse/IGNITE-17986?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Mirza Aliev updated IGNITE-17986:
---------------------------------
Description:
While we unmuted tests from `ItIgniteNodeRestartTest`, we noticed that restart
of the node seems to be broken.
It seems, that after https://issues.apache.org/jira/browse/IGNITE-17702 was
merged, some scenarios for a node recovery was broken. The root cause is highly
likely connected to the fact, that in the {{SchemaManager#start}} we directly
pass 0 to {{createSchema(0, tblId, tblName, desc).join();}}, which is
incorrect. Also {{registriesVv.complete(0);}} seems to be incorrect.
To reproduce the issue, run
{{ItIgniteNodeRestartTest#testTwoNodesRestartDirect}} or
{{ItIgniteInMemoryNodeRestartTest#inMemoryNodeRestartNotLeader}}
and we can see that such lines are presented in the log:
{noformat}
Caused by: java.lang.AssertionError: Token must be greater than actual
[token=0, actual=1]
at
org.apache.ignite.internal.causality.VersionedValue.checkToken(VersionedValue.java:597)
{noformat}
or
{noformat}
Caused by: java.lang.AssertionError: New token should be greater than current
[current=1, new=1]
at
org.apache.ignite.internal.causality.VersionedValue.completeOnRevision(VersionedValue.java:481)
{noformat}
was:
While we unmuted tests from `ItIgniteNodeRestartTest`, we noticed that restart
of the node seems to be broken.
It seems, that after https://issues.apache.org/jira/browse/IGNITE-17702 was
merged, some scenarios for a node recovery was broken. The root cause is highly
likely connected to the fact, that in the {{SchemaManager#start}} we directly
pass 0 to `createSchema(0, tblId, tblName, desc).join();`, which is incorrect.
Also {{registriesVv.complete(1);}} seems to be incorrect.
To reproduce the issue, run
{{ItIgniteNodeRestartTest#testTwoNodesRestartDirect}} or
{{ItIgniteInMemoryNodeRestartTest#inMemoryNodeRestartNotLeader}}
and we can see that such lines are presented in the log:
{noformat}
Caused by: java.lang.AssertionError: Token must be greater than actual
[token=0, actual=1]
at
org.apache.ignite.internal.causality.VersionedValue.checkToken(VersionedValue.java:597)
{noformat}
or
{noformat}
Caused by: java.lang.AssertionError: New token should be greater than current
[current=1, new=1]
at
org.apache.ignite.internal.causality.VersionedValue.completeOnRevision(VersionedValue.java:481)
{noformat}
> SchemaManager recovery was broken after refactoring
> ---------------------------------------------------
>
> Key: IGNITE-17986
> URL: https://issues.apache.org/jira/browse/IGNITE-17986
> Project: Ignite
> Issue Type: Bug
> Components: sql
> Affects Versions: 3.0.0-alpha5
> Reporter: Mirza Aliev
> Assignee: Evgeny Stanilovsky
> Priority: Major
> Labels: ignite-3
> Fix For: 3.0.0-beta2
>
>
> While we unmuted tests from `ItIgniteNodeRestartTest`, we noticed that
> restart of the node seems to be broken.
> It seems, that after https://issues.apache.org/jira/browse/IGNITE-17702 was
> merged, some scenarios for a node recovery was broken. The root cause is
> highly likely connected to the fact, that in the {{SchemaManager#start}} we
> directly pass 0 to {{createSchema(0, tblId, tblName, desc).join();}}, which
> is incorrect. Also {{registriesVv.complete(0);}} seems to be incorrect.
> To reproduce the issue, run
> {{ItIgniteNodeRestartTest#testTwoNodesRestartDirect}} or
> {{ItIgniteInMemoryNodeRestartTest#inMemoryNodeRestartNotLeader}}
> and we can see that such lines are presented in the log:
> {noformat}
> Caused by: java.lang.AssertionError: Token must be greater than actual
> [token=0, actual=1]
> at
> org.apache.ignite.internal.causality.VersionedValue.checkToken(VersionedValue.java:597)
> {noformat}
> or
> {noformat}
> Caused by: java.lang.AssertionError: New token should be greater than current
> [current=1, new=1]
> at
> org.apache.ignite.internal.causality.VersionedValue.completeOnRevision(VersionedValue.java:481)
> {noformat}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
