[
https://issues.apache.org/jira/browse/IGNITE-621?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Alexey Goncharuk updated IGNITE-621:
------------------------------------
Description:
Currently atomic cache throws a partial update exception when primary node
leave before response is received. Transactional cache throws topology
exception when primary node leaves grid before lock/prepare response is
received.
We can automatically retry update in the following cases:
# Unconditional cache operation (put, remove)
# Idempotent conditional cache operations (putIfAbsent, replace) can be
reinforced if value from primary node is forcibly pushed to backups regardless
whether update took place or not
# For entry processors repeat should be done if cache configuration flag
(should be introduced, defaults to false) is set to true.
# For transactional cache all implicit operations can be retried.
was:
Currently atomic cache throws a partial update exception when primary node
leave before response is received.
We can automatically retry update in the following cases:
# Unconditional cache operation (put, remove)
# Idempotent conditional cache operations (putIfAbsent, replace) can be
reinforced if value from primary node is forcibly pushed to backups regardless
whether update took place or not
# For entry processors repeat should be done if cache configuration flag
(should be introduced, defaults to false) is set to true.
> Need to add automatic retry to avoid partial update exception and topology
> exception
> ------------------------------------------------------------------------------------
>
> Key: IGNITE-621
> URL: https://issues.apache.org/jira/browse/IGNITE-621
> Project: Ignite
> Issue Type: Improvement
> Components: cache
> Affects Versions: sprint-2
> Reporter: Alexey Goncharuk
>
> Currently atomic cache throws a partial update exception when primary node
> leave before response is received. Transactional cache throws topology
> exception when primary node leaves grid before lock/prepare response is
> received.
> We can automatically retry update in the following cases:
> # Unconditional cache operation (put, remove)
> # Idempotent conditional cache operations (putIfAbsent, replace) can be
> reinforced if value from primary node is forcibly pushed to backups
> regardless whether update took place or not
> # For entry processors repeat should be done if cache configuration flag
> (should be introduced, defaults to false) is set to true.
> # For transactional cache all implicit operations can be retried.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
