[
https://issues.apache.org/jira/browse/IGNITE-27725?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Mirza Aliev updated IGNITE-27725:
---------------------------------
Release Note: Extended log coverage for Placement Driver mechanism to track
some edge cases during its recovery.
Description:
h3. Motivation
We would like to add some logs to Placement Driver and around, so we could
properly catch rare case when Placement Driver wasn't registered after
metastorage leader re-election and random nodes restarts.
* We have added logs to the most suspicious place --
{{LogicalTopologyEventListener#onTopologyLeap}}. This method wasn't even
implemented, meaning that if some node is lagging in terms of CMG raft group,
and rejoins after some time, it won't receive
{{LogicalTopologyEventListener#onNodeJoined}}, but {{onTopologyLeap}}, hence
all logic with node subscription will be missed. The first step is to log the
place and it must be fixed as soon as possible
https://issues.apache.org/jira/browse/IGNITE-27770
* Proper logging of {{LeaseTracker}} start and stop
* Starting Ragt client in {{PlacementDriverManager}}, receiving leader change
event.
etc.
was:
h3. Motivation
We would like to add some logs to Placement Driver and around, so we could
properly catch rare case when Placement Driver wasn't registered after
metastorage leader re-election and random nodes restarts.
* We have added logs to the most suspicious place in
{{TopologyAwareRaftGroupService}} --
{{LogicalTopologyEventListener#onTopologyLeap}} handler. This method wasn't
even implemented in {{TopologyAwareRaftGroupService}}, meaning that if some
node is lagging in terms of CMG raft group, and rejoins after some time, it
won't receive {{LogicalTopologyEventListener#onNodeJoined}}, but
{{onTopologyLeap}}, hence all logic with node subscription will be missed. The
first step is to log the place and it must be fixed as soon as possible
https://issues.apache.org/jira/browse/IGNITE-27770
* Proper logging of {{LeaseTracker}} start and stop
* Starting Raft client in {{PlacementDriverManager}}, receiving leader change
event.
etc.
> Improve placement driver log coverage
> -------------------------------------
>
> Key: IGNITE-27725
> URL: https://issues.apache.org/jira/browse/IGNITE-27725
> Project: Ignite
> Issue Type: Improvement
> Reporter: Mirza Aliev
> Assignee: Mirza Aliev
> Priority: Major
> Labels: ignite-3
> Fix For: 3.2
>
> Time Spent: 20m
> Remaining Estimate: 0h
>
> h3. Motivation
> We would like to add some logs to Placement Driver and around, so we could
> properly catch rare case when Placement Driver wasn't registered after
> metastorage leader re-election and random nodes restarts.
> * We have added logs to the most suspicious place --
> {{LogicalTopologyEventListener#onTopologyLeap}}. This method wasn't even
> implemented, meaning that if some node is lagging in terms of CMG raft group,
> and rejoins after some time, it won't receive
> {{LogicalTopologyEventListener#onNodeJoined}}, but {{onTopologyLeap}}, hence
> all logic with node subscription will be missed. The first step is to log the
> place and it must be fixed as soon as possible
> https://issues.apache.org/jira/browse/IGNITE-27770
> * Proper logging of {{LeaseTracker}} start and stop
> * Starting Ragt client in {{PlacementDriverManager}}, receiving leader change
> event.
> etc.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
