[JBoss JIRA] (ISPN-11474) RecoveryManager API should be non blocking
by Tristan Tarrant (Jira)
[ https://issues.redhat.com/browse/ISPN-11474?page=com.atlassian.jira.plugi... ]
Tristan Tarrant updated ISPN-11474:
-----------------------------------
Fix Version/s: 11.0.0.Dev04
(was: 11.0.0.Dev03)
> RecoveryManager API should be non blocking
> ------------------------------------------
>
> Key: ISPN-11474
> URL: https://issues.redhat.com/browse/ISPN-11474
> Project: Infinispan
> Issue Type: Sub-task
> Reporter: Will Burns
> Assignee: Will Burns
> Priority: Major
> Fix For: 11.0.0.Dev04
>
>
> RecoverManager, which is internal only, has a blocking API which can pass a sync parameter. Instead we should change the API to be non blocking and let callers decide if they want to wait or not on it. This allows us to clean up callers who may be able to use this in a non blocking fashion.
--
This message was sent by Atlassian Jira
(v7.13.8#713008)
6 years
[JBoss JIRA] (ISPN-11480) GlobalComponentRegistry shutdownHook should call EmbeddedCacheManager#stop
by Tristan Tarrant (Jira)
[ https://issues.redhat.com/browse/ISPN-11480?page=com.atlassian.jira.plugi... ]
Tristan Tarrant updated ISPN-11480:
-----------------------------------
Fix Version/s: 11.0.0.Dev04
(was: 11.0.0.Dev03)
> GlobalComponentRegistry shutdownHook should call EmbeddedCacheManager#stop
> --------------------------------------------------------------------------
>
> Key: ISPN-11480
> URL: https://issues.redhat.com/browse/ISPN-11480
> Project: Infinispan
> Issue Type: Bug
> Components: Core
> Affects Versions: 11.0.0.Alpha2, 10.1.5.Final
> Reporter: Ryan Emerson
> Assignee: Ryan Emerson
> Priority: Major
> Fix For: 11.0.0.Dev04, 10.1.6.Final
>
>
> Currently if a SIGTERM is called on a Infinispan processor, the shutdown hook calls {{GlobalComponentRegistry::stop}} which results in caches not leaving the cluster gracefully. In particular this affects the behaviour of 2 node Infinispan clusters when utlised in a Kubernetes StatefulSet, as it will result in the cluster becoming degraded during a rolling upgrade.
--
This message was sent by Atlassian Jira
(v7.13.8#713008)
6 years
[JBoss JIRA] (ISPN-11489) TransactionCoordinator updated for non blocking
by Tristan Tarrant (Jira)
[ https://issues.redhat.com/browse/ISPN-11489?page=com.atlassian.jira.plugi... ]
Tristan Tarrant updated ISPN-11489:
-----------------------------------
Fix Version/s: 11.0.0.Dev04
(was: 11.0.0.Dev03)
> TransactionCoordinator updated for non blocking
> -----------------------------------------------
>
> Key: ISPN-11489
> URL: https://issues.redhat.com/browse/ISPN-11489
> Project: Infinispan
> Issue Type: Sub-task
> Reporter: Will Burns
> Assignee: Will Burns
> Priority: Major
> Fix For: 11.0.0.Dev04
>
>
> TransactionCoordinator is utilized when working with transactions. Unfortunately due to transaction API being inherently blocking, we need to make some exceptions to how this is handled.
> # TransactionCoordinator should only be invokable by a blocking method, the inverse of the PersistenceManagerImpl
> # We should invoke all the non blocking code of TransactionCoordinator in a non blocking thread and then block on its completion in the blocking thread.
--
This message was sent by Atlassian Jira
(v7.13.8#713008)
6 years
[JBoss JIRA] (ISPN-11465) Upgrade to JGroups 4.2.1.Final
by Tristan Tarrant (Jira)
[ https://issues.redhat.com/browse/ISPN-11465?page=com.atlassian.jira.plugi... ]
Tristan Tarrant updated ISPN-11465:
-----------------------------------
Fix Version/s: 11.0.0.Dev04
(was: 11.0.0.Dev03)
> Upgrade to JGroups 4.2.1.Final
> ------------------------------
>
> Key: ISPN-11465
> URL: https://issues.redhat.com/browse/ISPN-11465
> Project: Infinispan
> Issue Type: Component Upgrade
> Components: Dependency
> Affects Versions: 11.0.0.Alpha2
> Reporter: Dan Berindei
> Assignee: Dan Berindei
> Priority: Major
> Fix For: 11.0.0.Dev04
>
>
> JGroups 4.2.1 includes a fix for JGRP-2435, which should fix the random failure in {{InitialClusterSizeTest}}.
> JGRP-2451 also adds a new failure detection protocol, {{FD3_ALL}}, which doesn't need extra heartbeats as long as the nodes are sending other broadcast messages.
--
This message was sent by Atlassian Jira
(v7.13.8#713008)
6 years