[
https://issues.apache.org/jira/browse/CLOUDSTACK-6740?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Koushik Das updated CLOUDSTACK-6740:
------------------------------------
Description:
List of changes:
1. Create a separate thread pool for handling cron tasks. Currently normal and
cron tasks are all handled by the same direct agent thread pool.
2. For normal tasks (generated as part of user/admin API calls), if throttle
limit reached then tasks will be queued up for subsequent execution. This will
avoid uncertainty as to when a certain command pass/fail based on the
throttling. For now the queue is unbounded but this needs to be revisited. One
possibility is to change the agent state to something appropriate, based on the
queue limit, so that commands are not sent to this host from the upper layers
of MS.
3. For cron tasks (internally generated by MS like ping, VM sync etc.), if
throttle limit reached then it will be rejected. Since these are internally
generated these can be rejected without any issues.
was:
List of changes:
1. Create a separate thread pool for handling cron tasks. Currently normal and
cron tasks are all handled by the same direct agent thread pool.
2. For normal tasks (generated as part of user/admin API calls), if throttle
limit reached then tasks will be queued up for subsequent execution. This will
avoid uncertainty as to when a certain command pass/fail based on the
throttling. For now the queue is unbounded but this needs to be revisited. One
possibility is to change the agent state to something appropriate, based on the
queue limit, so that commands are sent to this host from the upper layers of MS.
3. For cron tasks (internally generated by MS like ping, VM sync etc.), if
throttle limit reached then it will be rejected. Since these are internally
generated these can be rejected without any issues.
> Direct agent command throttling improvements
> --------------------------------------------
>
> Key: CLOUDSTACK-6740
> URL: https://issues.apache.org/jira/browse/CLOUDSTACK-6740
> Project: CloudStack
> Issue Type: Bug
> Security Level: Public(Anyone can view this level - this is the
> default.)
> Components: Management Server
> Affects Versions: 4.5.0
> Reporter: Koushik Das
> Assignee: Koushik Das
> Fix For: 4.5.0
>
>
> List of changes:
> 1. Create a separate thread pool for handling cron tasks. Currently normal
> and cron tasks are all handled by the same direct agent thread pool.
> 2. For normal tasks (generated as part of user/admin API calls), if throttle
> limit reached then tasks will be queued up for subsequent execution. This
> will avoid uncertainty as to when a certain command pass/fail based on the
> throttling. For now the queue is unbounded but this needs to be revisited.
> One possibility is to change the agent state to something appropriate, based
> on the queue limit, so that commands are not sent to this host from the upper
> layers of MS.
> 3. For cron tasks (internally generated by MS like ping, VM sync etc.), if
> throttle limit reached then it will be rejected. Since these are internally
> generated these can be rejected without any issues.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
