[jira] [Updated] (SOLR-10643) Throttling strategy for triggers and policy executions
[ https://issues.apache.org/jira/browse/SOLR-10643?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Shalin Shekhar Mangar updated SOLR-10643: - Fix Version/s: (was: 7.0) 7.1 master (8.0) > Throttling strategy for triggers and policy executions > -- > > Key: SOLR-10643 > URL: https://issues.apache.org/jira/browse/SOLR-10643 > Project: Solr > Issue Type: Sub-task > Security Level: Public(Default Security Level. Issues are Public) > Components: SolrCloud >Reporter: Shalin Shekhar Mangar >Assignee: Shalin Shekhar Mangar > Labels: autoscaling > Fix For: master (8.0), 7.1 > > Attachments: SOLR-10643.patch, SOLR-10643.patch, SOLR-10643.patch > > > We must ensure that triggers and policy executions: > # Do not step on each other's toes by concurrent executions > # Do not fire/execute too frequently > # Do not stack up -- This message was sent by Atlassian JIRA (v6.4.14#64029) - To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org
[jira] [Updated] (SOLR-10643) Throttling strategy for triggers and policy executions
[ https://issues.apache.org/jira/browse/SOLR-10643?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Shalin Shekhar Mangar updated SOLR-10643: - Attachment: SOLR-10643.patch Added a sanity check against concurrent execution of actions. This is ready. I'll commit it to the autoscaling branch. > Throttling strategy for triggers and policy executions > -- > > Key: SOLR-10643 > URL: https://issues.apache.org/jira/browse/SOLR-10643 > Project: Solr > Issue Type: Sub-task > Security Level: Public(Default Security Level. Issues are Public) > Components: SolrCloud >Reporter: Shalin Shekhar Mangar >Assignee: Shalin Shekhar Mangar > Labels: autoscaling > Fix For: master (7.0) > > Attachments: SOLR-10643.patch, SOLR-10643.patch, SOLR-10643.patch > > > We must ensure that triggers and policy executions: > # Do not step on each other's toes by concurrent executions > # Do not fire/execute too frequently > # Do not stack up -- This message was sent by Atlassian JIRA (v6.3.15#6346) - To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org
[jira] [Updated] (SOLR-10643) Throttling strategy for triggers and policy executions
[ https://issues.apache.org/jira/browse/SOLR-10643?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Shalin Shekhar Mangar updated SOLR-10643: - Attachment: SOLR-10643.patch This patch addresses problem #2. Changes: # Introduced an action throttle with a default wait of 5 seconds -- it'd be nice to configure this wait time but currently there is no way of setting such properties. I'll open another issue for a set-properties autoscaling api. # Added TriggerIntegrationTest.testTriggerThrottling for testing throttling # Added NodeLostTriggerTest.testListenerAcceptance for testing that state is maintained if listener is not ready > Throttling strategy for triggers and policy executions > -- > > Key: SOLR-10643 > URL: https://issues.apache.org/jira/browse/SOLR-10643 > Project: Solr > Issue Type: Sub-task > Security Level: Public(Default Security Level. Issues are Public) > Components: SolrCloud >Reporter: Shalin Shekhar Mangar >Assignee: Shalin Shekhar Mangar > Labels: autoscaling > Fix For: master (7.0) > > Attachments: SOLR-10643.patch, SOLR-10643.patch > > > We must ensure that triggers and policy executions: > # Do not step on each other's toes by concurrent executions > # Do not fire/execute too frequently > # Do not stack up -- This message was sent by Atlassian JIRA (v6.3.15#6346) - To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org
[jira] [Updated] (SOLR-10643) Throttling strategy for triggers and policy executions
[
https://issues.apache.org/jira/browse/SOLR-10643?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Shalin Shekhar Mangar updated SOLR-10643:
-
Attachment: SOLR-10643.patch
This patch adds protection against points #1 and #3.
# The TriggerListener.triggerFired method now returns a boolean. The return
value is true if the fire was ready to be processed by the listener otherwise
not.
# If the value returned by the above was false then triggers are responsible
for maintaining state such that the fire can be attempted again on subsequent
executions until the listener returns true.
# The ScheduledTriggers class prevents triggers to be run between the time a
previous trigger fire is submitted and it finishes execution
# In case multiple triggers were executing concurrently, it prevents trigger
fires from stacking up on the action executor.
# Added javadocs in relevant places to document this behavior and expectations
from subclasses of TriggerListener
# Added NodeAddedTriggerTest.testListenerAcceptance() to test this behavior
Todo
# Add similar test for NodeLostTrigger
# Add an integration test for this behavior
# Add rate limiting between action executions (point #2 in description)
Note that the {{waitFor}} implicitly specifies a cooling off period but for the
same trigger only. We want another cooling off period between different
triggers as well.
> Throttling strategy for triggers and policy executions
> --
>
> Key: SOLR-10643
> URL: https://issues.apache.org/jira/browse/SOLR-10643
> Project: Solr
> Issue Type: Sub-task
> Security Level: Public(Default Security Level. Issues are Public)
> Components: SolrCloud
>Reporter: Shalin Shekhar Mangar
>Assignee: Shalin Shekhar Mangar
> Labels: autoscaling
> Fix For: master (7.0)
>
> Attachments: SOLR-10643.patch
>
>
> We must ensure that triggers and policy executions:
> # Do not step on each other's toes by concurrent executions
> # Do not fire/execute too frequently
> # Do not stack up
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
-
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
