You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@ignite.apache.org by "Yakov Zhdanov (JIRA)" <ji...@apache.org> on 2016/12/05 08:34:58 UTC
[jira] [Updated] (IGNITE-75) Revisit node stop procedure and busy
locking
[ https://issues.apache.org/jira/browse/IGNITE-75?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Yakov Zhdanov updated IGNITE-75:
--------------------------------
Assignee: Yakov Zhdanov
Description:
I think we need to eliminate all the gateways and busy locks in order to reduce inter-thread operations and get closer to share nothing pattern.
Here is the original description (not sure if it still makes sense):
# Have 64 or 128 threads preforming operations from public API (e.g. cache.put(..))
# Have 1 thread to stop current node. Stop will most likely hang due to busy lock on kernal gateway (since it is busy locked by threads performing puts)
Comments:
# Why we have multiple busy locks instead of using one from KernalGateway?
# All ongoing user initiated operations should fail, no new allowed and then stop procedure should proceed.
# One can use {{GridCacheStopSelfTest}} to debug this, but create a new test which should involve 2 or more nodes.
was:
# Have 64 or 128 threads preforming operations from public API (e.g. cache.put(..))
# Have 1 thread to stop current node. Stop will most likely hang due to busy lock on kernal gateway (since it is busy locked by threads performing puts)
Comments:
# Why we have multiple busy locks instead of using one from KernalGateway?
# All ongoing user initiated operations should fail, no new allowed and then stop procedure should proceed.
# One can use {{GridCacheStopSelfTest}} to debug this, but create a new test which should involve 2 or more nodes.
Issue Type: Improvement (was: Bug)
> Revisit node stop procedure and busy locking
> --------------------------------------------
>
> Key: IGNITE-75
> URL: https://issues.apache.org/jira/browse/IGNITE-75
> Project: Ignite
> Issue Type: Improvement
> Reporter: Yakov Zhdanov
> Assignee: Yakov Zhdanov
>
> I think we need to eliminate all the gateways and busy locks in order to reduce inter-thread operations and get closer to share nothing pattern.
> Here is the original description (not sure if it still makes sense):
> # Have 64 or 128 threads preforming operations from public API (e.g. cache.put(..))
> # Have 1 thread to stop current node. Stop will most likely hang due to busy lock on kernal gateway (since it is busy locked by threads performing puts)
> Comments:
> # Why we have multiple busy locks instead of using one from KernalGateway?
> # All ongoing user initiated operations should fail, no new allowed and then stop procedure should proceed.
> # One can use {{GridCacheStopSelfTest}} to debug this, but create a new test which should involve 2 or more nodes.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)