[
https://issues.apache.org/jira/browse/IGNITE-20397?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Sergey Uttsel updated IGNITE-20397:
-----------------------------------
Description:
h3. Motivation
{code:java}
java.lang.AssertionError: Group of the event is unsupported
[nodeId=<11_part_18/isaat_n_2>,
event=org.apache.ignite.raft.jraft.core.NodeImpl$LogEntryAndClosure@653d84a]
at
org.apache.ignite.raft.jraft.disruptor.StripedDisruptor$StripeEntryHandler.onEvent(StripedDisruptor.java:224)
~[ignite-raft-3.0.0-SNAPSHOT.jar:?]
at
org.apache.ignite.raft.jraft.disruptor.StripedDisruptor$StripeEntryHandler.onEvent(StripedDisruptor.java:191)
~[ignite-raft-3.0.0-SNAPSHOT.jar:?]
at com.lmax.disruptor.BatchEventProcessor.run(BatchEventProcessor.java:137)
~[disruptor-3.3.7.jar:?]
at java.lang.Thread.run(Thread.java:834) ~[?:?] {code}
[https://ci.ignite.apache.org/buildConfiguration/ApacheIgnite3xGradle_Test_RunAllTests/7498320?expandCode+Inspection=true&expandBuildProblemsSection=true&hideProblemsFromDependencies=false&expandBuildTestsSection=true&hideTestsFromDependencies=false&expandBuildChangesSection=true]
The root cause:
# StripedDisruptor.StripeEntryHandler#onEvent method gets handler from
StripedDisruptor.StripeEntryHandler#subscribers by event.nodeId().
# In some cases the `subscribers` map is cleared by invocation of
StripedDisruptor.StripeEntryHandler#unsubscribe (for example on table
dropping), and then StripeEntryHandler receives event with
SafeTimeSyncCommandImpl.
# It produces an assertion error: `assert handler != null`
The issue is not caused by the catalog feature changes.
The issue is reproduced when I run the ItSqlAsynchronousApiTest#batchIncomplete
with RepeatedTest annotation. In this case the cluster is not restarted after
each tests. It possible to reproduced it frequently if add Thread.sleep in
StripeEntryHandler#onEvent.
h3. Implementation notes
We decided that we can use LOG.warn() instead of an assert because it is safely
to skip this event if the table was dropped.
{code:java}
if (handler != null) {
handler.onEvent(event, sequence, endOfBatch || subscribers.size() > 1 &&
!supportsBatches);
} else {
LOG.warn(format("Group of the event is unsupported [nodeId={}, event={}]",
event.nodeId(), event));
} {code}
It is temp solution and we need to add TODO with link
https://issues.apache.org/jira/browse/IGNITE-20536
*Definition of done*
There is no asserts if handler is null.
was:
h3. Motivation
{code:java}
java.lang.AssertionError: Group of the event is unsupported
[nodeId=<11_part_18/isaat_n_2>,
event=org.apache.ignite.raft.jraft.core.NodeImpl$LogEntryAndClosure@653d84a]
at
org.apache.ignite.raft.jraft.disruptor.StripedDisruptor$StripeEntryHandler.onEvent(StripedDisruptor.java:224)
~[ignite-raft-3.0.0-SNAPSHOT.jar:?]
at
org.apache.ignite.raft.jraft.disruptor.StripedDisruptor$StripeEntryHandler.onEvent(StripedDisruptor.java:191)
~[ignite-raft-3.0.0-SNAPSHOT.jar:?]
at com.lmax.disruptor.BatchEventProcessor.run(BatchEventProcessor.java:137)
~[disruptor-3.3.7.jar:?]
at java.lang.Thread.run(Thread.java:834) ~[?:?] {code}
[https://ci.ignite.apache.org/buildConfiguration/ApacheIgnite3xGradle_Test_RunAllTests/7498320?expandCode+Inspection=true&expandBuildProblemsSection=true&hideProblemsFromDependencies=false&expandBuildTestsSection=true&hideTestsFromDependencies=false&expandBuildChangesSection=true]
The root cause:
# StripedDisruptor.StripeEntryHandler#onEvent method gets handler from
StripedDisruptor.StripeEntryHandler#subscribers by event.nodeId().
# In some cases the `subscribers` map is cleared by invocation of
StripedDisruptor.StripeEntryHandler#unsubscribe (for example on table
dropping), and then StripeEntryHandler receives event with
SafeTimeSyncCommandImpl.
# It produces an assertion error: `assert handler != null`
The issue is not caused by the catalog feature changes.
The issue is reproduced when I run the ItSqlAsynchronousApiTest#batchIncomplete
with RepeatedTest annotation. In this case the cluster is not restarted after
each tests. It possible to reproduced it frequently if add Thread.sleep in
StripeEntryHandler#onEvent.
h3. Implementation notes
We decided that we can use LOG.warn() instead of an assert because it is safely
to skip this event if the table was dropped.
{code:java}
if (handler != null) {
handler.onEvent(event, sequence, endOfBatch || subscribers.size() > 1 &&
!supportsBatches);
} else {
LOG.warn(format("Group of the event is unsupported [nodeId={}, event={}]",
event.nodeId(), event));
} {code}
*Definition of done*
There is no asserts if handler is null.
> java.lang.AssertionError: Group of the event is unsupported
> -----------------------------------------------------------
>
> Key: IGNITE-20397
> URL: https://issues.apache.org/jira/browse/IGNITE-20397
> Project: Ignite
> Issue Type: Bug
> Reporter: Alexander Lapin
> Priority: Major
> Labels: ignite-3
> Time Spent: 10m
> Remaining Estimate: 0h
>
> h3. Motivation
> {code:java}
> java.lang.AssertionError: Group of the event is unsupported
> [nodeId=<11_part_18/isaat_n_2>,
> event=org.apache.ignite.raft.jraft.core.NodeImpl$LogEntryAndClosure@653d84a]
> at
> org.apache.ignite.raft.jraft.disruptor.StripedDisruptor$StripeEntryHandler.onEvent(StripedDisruptor.java:224)
> ~[ignite-raft-3.0.0-SNAPSHOT.jar:?]
> at
> org.apache.ignite.raft.jraft.disruptor.StripedDisruptor$StripeEntryHandler.onEvent(StripedDisruptor.java:191)
> ~[ignite-raft-3.0.0-SNAPSHOT.jar:?]
> at
> com.lmax.disruptor.BatchEventProcessor.run(BatchEventProcessor.java:137)
> ~[disruptor-3.3.7.jar:?]
> at java.lang.Thread.run(Thread.java:834) ~[?:?] {code}
> [https://ci.ignite.apache.org/buildConfiguration/ApacheIgnite3xGradle_Test_RunAllTests/7498320?expandCode+Inspection=true&expandBuildProblemsSection=true&hideProblemsFromDependencies=false&expandBuildTestsSection=true&hideTestsFromDependencies=false&expandBuildChangesSection=true]
> The root cause:
> # StripedDisruptor.StripeEntryHandler#onEvent method gets handler from
> StripedDisruptor.StripeEntryHandler#subscribers by event.nodeId().
> # In some cases the `subscribers` map is cleared by invocation of
> StripedDisruptor.StripeEntryHandler#unsubscribe (for example on table
> dropping), and then StripeEntryHandler receives event with
> SafeTimeSyncCommandImpl.
> # It produces an assertion error: `assert handler != null`
> The issue is not caused by the catalog feature changes.
> The issue is reproduced when I run the
> ItSqlAsynchronousApiTest#batchIncomplete with RepeatedTest annotation. In
> this case the cluster is not restarted after each tests. It possible to
> reproduced it frequently if add Thread.sleep in StripeEntryHandler#onEvent.
> h3. Implementation notes
> We decided that we can use LOG.warn() instead of an assert because it is
> safely to skip this event if the table was dropped.
> {code:java}
> if (handler != null) {
> handler.onEvent(event, sequence, endOfBatch || subscribers.size() > 1 &&
> !supportsBatches);
> } else {
> LOG.warn(format("Group of the event is unsupported [nodeId={},
> event={}]", event.nodeId(), event));
> } {code}
> It is temp solution and we need to add TODO with link
> https://issues.apache.org/jira/browse/IGNITE-20536
> *Definition of done*
> There is no asserts if handler is null.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
