[
https://issues.apache.org/jira/browse/IGNITE-4265?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Yakov Zhdanov updated IGNITE-4265:
----------------------------------
Description:
I. Shared fs checkpoint SPI seems to be totally wrong
Problems spotted in code
# Blocking IO API
# No locks or synchronization
Please see GridSharedFsCheckpointSpiMultiThreadedSelfTest
I would think of removing shared FS checkpoints and move to distributed cache.
II. Having the above I would do:
# Revisit checkpoints SPI/manager. I would probably replace it with checkoints
processor using distributed cache(s)
# distributed caches can be configured in configuration or on the fly using
templates. This needs to be elaborated further.
#
org.apache.ignite.internal.managers.checkpoint.GridCheckpointManager#onSessionEnd
seems to be handy for clean up and can be implemented with cache queries -
session ID should be added to each entry
was:
I. Shared fs checkpoint SPI seems to be totally wrong
Problems spotted in code
# Blocking IO API
# No locks or synchronization
Please see GridSharedFsCheckpointSpiMultiThreadedSelfTest
I would think of removing shared FS checkpoints and move to distributed cache.
II. Having the above I would do:
# Revisit checkpoints SPI/manager. I would probably replace it with checkoints
processor using distributed cache(s)
# distributed caches can be configured in configuration or on the fly using
templates. This needs to be elaborated further.
#
org.apache.ignite.internal.managers.checkpoint.GridCheckpointManager#onSessionEnd
seems to be handy for clean up and can be implemented with cache queries -
session ID should be added to each entry
#
> Need to revisit compute grid checkpoints
> ----------------------------------------
>
> Key: IGNITE-4265
> URL: https://issues.apache.org/jira/browse/IGNITE-4265
> Project: Ignite
> Issue Type: Task
> Reporter: Yakov Zhdanov
> Priority: Minor
>
> I. Shared fs checkpoint SPI seems to be totally wrong
> Problems spotted in code
> # Blocking IO API
> # No locks or synchronization
> Please see GridSharedFsCheckpointSpiMultiThreadedSelfTest
> I would think of removing shared FS checkpoints and move to distributed cache.
> II. Having the above I would do:
> # Revisit checkpoints SPI/manager. I would probably replace it with
> checkoints processor using distributed cache(s)
> # distributed caches can be configured in configuration or on the fly using
> templates. This needs to be elaborated further.
> #
> org.apache.ignite.internal.managers.checkpoint.GridCheckpointManager#onSessionEnd
> seems to be handy for clean up and can be implemented with cache queries -
> session ID should be added to each entry
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
