[
https://issues.apache.org/jira/browse/IGNITE-27203?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Vladislav Pyatkov updated IGNITE-27203:
---------------------------------------
Description:
h3. Motivation
Operations on user data are often time-limited. In most cases, such operations
are components of a distributed transaction, which itself has an upper timeout
limit.
Accordingly, all replication commands on partition groups have to be
time-limited (currently a default timeout is used).
At the same time, internal system operations (such as changing assignments or
crash recovery) require waiting for the replication group’s response for an
effectively unlimited duration (we currently have a WA IGNITE-23633).
h3. Implementation notes
Consider the approach using a topology-aware RAFT client (IGNITE-27156).
h3. Definition of done
Ability to wait for replication operations for an effectively infinite amount
of time.
was:
h3. Motivation
Operations on user data are often time-limited. In most cases, such operations
are components of a distributed transaction, which itself has an upper timeout
limit.
Accordingly, all replication commands on partition groups must be time-limited
(currently a default timeout is used).
At the same time, internal system operations (such as changing assignments or
crash recovery) require waiting for the replication group’s response for an
effectively unlimited duration (we currently have a WA IGNITE-23633).
h3. Implementation notes
Consider the approach using a topology-aware RAFT client (IGNITE-27156).
h3. Definition of done
Ability to wait for replication operations for an effectively infinite amount
of time.
> Support waiting for RAFT client of partitions
> ---------------------------------------------
>
> Key: IGNITE-27203
> URL: https://issues.apache.org/jira/browse/IGNITE-27203
> Project: Ignite
> Issue Type: Improvement
> Components: replication ai3
> Reporter: Vladislav Pyatkov
> Priority: Major
> Labels: ignite-3, tech-debt
>
> h3. Motivation
> Operations on user data are often time-limited. In most cases, such
> operations are components of a distributed transaction, which itself has an
> upper timeout limit.
> Accordingly, all replication commands on partition groups have to be
> time-limited (currently a default timeout is used).
> At the same time, internal system operations (such as changing assignments or
> crash recovery) require waiting for the replication group’s response for an
> effectively unlimited duration (we currently have a WA IGNITE-23633).
> h3. Implementation notes
> Consider the approach using a topology-aware RAFT client (IGNITE-27156).
> h3. Definition of done
> Ability to wait for replication operations for an effectively infinite amount
> of time.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
