[jira] [Commented] (SOLR-12463) Search rate trigger should check metrics less often

[Varun Thacker (JIRA)]

[ 
https://issues.apache.org/jira/browse/SOLR-12463?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16505133#comment-16505133
 ] 

Varun Thacker commented on SOLR-12463:
--

Yeah 10 sounds fine to me.  I'll file two separate Jiras and close this one 
out. 

> Search rate trigger should check metrics less often
> ---
>
> Key: SOLR-12463
> URL: https://issues.apache.org/jira/browse/SOLR-12463
> Project: Solr
>  Issue Type: Bug
>  Security Level: Public(Default Security Level. Issues are Public) 
>  Components: AutoScaling
>Reporter: Varun Thacker
>Priority: Major
>
> I created a search rate trigger by running this command 
> {code:java}
> curl -X POST -H 'Content-type:application/json' --data-binary '{
> "set-trigger": {
> "name" : "search_rate_trigger",
> "event" : "searchRate",
> "collection" : "gettingstarted",
> "rate" : 1.0,
> "waitFor" : "1m",
> "enabled" : true,
> "actions" : [
> {
> "name" : "compute_plan",
> "class": "solr.ComputePlanAction"
> },
> {
> "name" : "execute_plan",
> "class": "solr.ExecutePlanAction"
> }
> ]
> }
> }' http://localhost:8983/solr/admin/autoscaling
> {code}
> As soon as I add this by logs start getting spammed with these every 1 second
> {code:java}
> INFO  - 2018-06-07 15:29:16.243; [   ] org.apache.solr.servlet.HttpSolrCall; 
> [admin] webapp=null path=/admin/metrics 
> params={wt=javabin&version=2&key=solr.core.gettingstarted.shard2.replica_n4:QUERY./select.requestTimes:1minRate&key=solr.core.gettingstarted.shard1.replica_n1:QUERY./select.requestTimes:1minRate}
>  status=0 QTime=0
> INFO  - 2018-06-07 15:29:17.253; [   ] org.apache.solr.servlet.HttpSolrCall; 
> [admin] webapp=null path=/admin/metrics 
> params={wt=javabin&version=2&key=solr.core.gettingstarted.shard2.replica_n4:QUERY./select.requestTimes:1minRate&key=solr.core.gettingstarted.shard1.replica_n1:QUERY./select.requestTimes:1minRate}
>  status=0 QTime=0
> INFO  - 2018-06-07 15:29:18.263; [   ] org.apache.solr.servlet.HttpSolrCall; 
> [admin] webapp=null path=/admin/metrics 
> params={wt=javabin&version=2&key=solr.core.gettingstarted.shard2.replica_n4:QUERY./select.requestTimes:1minRate&key=solr.core.gettingstarted.shard1.replica_n1:QUERY./select.requestTimes:1minRate}
>  status=0 QTime=0
> {code}
> Does 1 second seem to aggressive ? 
> Maybe the waitFor could be taken into account and query the metrics at 
> waitFor/2 or something ? 



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

-
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org

[jira] [Commented] (SOLR-12463) Search rate trigger should check metrics less often

[Shalin Shekhar Mangar (JIRA)]

[ 
https://issues.apache.org/jira/browse/SOLR-12463?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16505125#comment-16505125
 ] 

Shalin Shekhar Mangar commented on SOLR-12463:
--

Setting null should reset to the default -- it is a bug.

bq. Also do you think the default of 1 second is too fine grained? Maybe 15 
seconds? Just thinking aloud 

yeah, once a second was a good default back when we had only node added and 
lost triggers. We should bump it up higher now, maybe 10 seconds is okay?

> Search rate trigger should check metrics less often
> ---
>
> Key: SOLR-12463
> URL: https://issues.apache.org/jira/browse/SOLR-12463
> Project: Solr
>  Issue Type: Bug
>  Security Level: Public(Default Security Level. Issues are Public) 
>  Components: AutoScaling
>Reporter: Varun Thacker
>Priority: Major
>
> I created a search rate trigger by running this command 
> {code:java}
> curl -X POST -H 'Content-type:application/json' --data-binary '{
> "set-trigger": {
> "name" : "search_rate_trigger",
> "event" : "searchRate",
> "collection" : "gettingstarted",
> "rate" : 1.0,
> "waitFor" : "1m",
> "enabled" : true,
> "actions" : [
> {
> "name" : "compute_plan",
> "class": "solr.ComputePlanAction"
> },
> {
> "name" : "execute_plan",
> "class": "solr.ExecutePlanAction"
> }
> ]
> }
> }' http://localhost:8983/solr/admin/autoscaling
> {code}
> As soon as I add this by logs start getting spammed with these every 1 second
> {code:java}
> INFO  - 2018-06-07 15:29:16.243; [   ] org.apache.solr.servlet.HttpSolrCall; 
> [admin] webapp=null path=/admin/metrics 
> params={wt=javabin&version=2&key=solr.core.gettingstarted.shard2.replica_n4:QUERY./select.requestTimes:1minRate&key=solr.core.gettingstarted.shard1.replica_n1:QUERY./select.requestTimes:1minRate}
>  status=0 QTime=0
> INFO  - 2018-06-07 15:29:17.253; [   ] org.apache.solr.servlet.HttpSolrCall; 
> [admin] webapp=null path=/admin/metrics 
> params={wt=javabin&version=2&key=solr.core.gettingstarted.shard2.replica_n4:QUERY./select.requestTimes:1minRate&key=solr.core.gettingstarted.shard1.replica_n1:QUERY./select.requestTimes:1minRate}
>  status=0 QTime=0
> INFO  - 2018-06-07 15:29:18.263; [   ] org.apache.solr.servlet.HttpSolrCall; 
> [admin] webapp=null path=/admin/metrics 
> params={wt=javabin&version=2&key=solr.core.gettingstarted.shard2.replica_n4:QUERY./select.requestTimes:1minRate&key=solr.core.gettingstarted.shard1.replica_n1:QUERY./select.requestTimes:1minRate}
>  status=0 QTime=0
> {code}
> Does 1 second seem to aggressive ? 
> Maybe the waitFor could be taken into account and query the metrics at 
> waitFor/2 or something ? 



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

-
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org

[jira] [Commented] (SOLR-12463) Search rate trigger should check metrics less often

[Varun Thacker (JIRA)]

[ 
https://issues.apache.org/jira/browse/SOLR-12463?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16505073#comment-16505073
 ] 

Varun Thacker commented on SOLR-12463:
--

Fair enough. I wasn't aware of this setting so this should work for me.

I noticed one thing that is maybe a bug though?

1. By default the metrics get queried every 1 second

2. Now I change it to 5 second via this API and this works
{code:java}
curl -X POST -H 'Content-type:application/json' --data-binary '{ 
"set-properties": { "triggerScheduleDelaySeconds": 1 } }' 
http://localhost:8983/solr/admin/autoscaling{code}
3. I want to change it back to the default, so I applied null but that didn't 
work ( the metrics is still getting queried at 5 second interval and not 1 
second).  Explicitly setting it to 1 works 
{code:java}
curl -X POST -H 'Content-type:application/json' --data-binary '{ 
"set-properties": { "triggerScheduleDelaySeconds": null } }' 
http://localhost:8983/solr/admin/autoscaling{code}
 

Also do you think the default of 1 second is too fine grained? Maybe 15 
seconds? Just thinking aloud 

> Search rate trigger should check metrics less often
> ---
>
> Key: SOLR-12463
> URL: https://issues.apache.org/jira/browse/SOLR-12463
> Project: Solr
>  Issue Type: Bug
>  Security Level: Public(Default Security Level. Issues are Public) 
>  Components: AutoScaling
>Reporter: Varun Thacker
>Priority: Major
>
> I created a search rate trigger by running this command 
> {code:java}
> curl -X POST -H 'Content-type:application/json' --data-binary '{
> "set-trigger": {
> "name" : "search_rate_trigger",
> "event" : "searchRate",
> "collection" : "gettingstarted",
> "rate" : 1.0,
> "waitFor" : "1m",
> "enabled" : true,
> "actions" : [
> {
> "name" : "compute_plan",
> "class": "solr.ComputePlanAction"
> },
> {
> "name" : "execute_plan",
> "class": "solr.ExecutePlanAction"
> }
> ]
> }
> }' http://localhost:8983/solr/admin/autoscaling
> {code}
> As soon as I add this by logs start getting spammed with these every 1 second
> {code:java}
> INFO  - 2018-06-07 15:29:16.243; [   ] org.apache.solr.servlet.HttpSolrCall; 
> [admin] webapp=null path=/admin/metrics 
> params={wt=javabin&version=2&key=solr.core.gettingstarted.shard2.replica_n4:QUERY./select.requestTimes:1minRate&key=solr.core.gettingstarted.shard1.replica_n1:QUERY./select.requestTimes:1minRate}
>  status=0 QTime=0
> INFO  - 2018-06-07 15:29:17.253; [   ] org.apache.solr.servlet.HttpSolrCall; 
> [admin] webapp=null path=/admin/metrics 
> params={wt=javabin&version=2&key=solr.core.gettingstarted.shard2.replica_n4:QUERY./select.requestTimes:1minRate&key=solr.core.gettingstarted.shard1.replica_n1:QUERY./select.requestTimes:1minRate}
>  status=0 QTime=0
> INFO  - 2018-06-07 15:29:18.263; [   ] org.apache.solr.servlet.HttpSolrCall; 
> [admin] webapp=null path=/admin/metrics 
> params={wt=javabin&version=2&key=solr.core.gettingstarted.shard2.replica_n4:QUERY./select.requestTimes:1minRate&key=solr.core.gettingstarted.shard1.replica_n1:QUERY./select.requestTimes:1minRate}
>  status=0 QTime=0
> {code}
> Does 1 second seem to aggressive ? 
> Maybe the waitFor could be taken into account and query the metrics at 
> waitFor/2 or something ? 



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

-
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org

[jira] [Commented] (SOLR-12463) Search rate trigger should check metrics less often

[Shalin Shekhar Mangar (JIRA)]

[ 
https://issues.apache.org/jira/browse/SOLR-12463?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16504967#comment-16504967
 ] 

Shalin Shekhar Mangar commented on SOLR-12463:
--

You can change the trigger delay by setting an autoscaling property called 
"triggerScheduleDelaySeconds". However, it changes the delay for all triggers. 
We don't support different delays for different triggers today. Doing so seems 
complicated to me without much upside.

> Search rate trigger should check metrics less often
> ---
>
> Key: SOLR-12463
> URL: https://issues.apache.org/jira/browse/SOLR-12463
> Project: Solr
>  Issue Type: Bug
>  Security Level: Public(Default Security Level. Issues are Public) 
>  Components: AutoScaling
>Reporter: Varun Thacker
>Priority: Major
>
> I created a search rate trigger by running this command 
> {code:java}
> curl -X POST -H 'Content-type:application/json' --data-binary '{
> "set-trigger": {
> "name" : "search_rate_trigger",
> "event" : "searchRate",
> "collection" : "gettingstarted",
> "rate" : 1.0,
> "waitFor" : "1m",
> "enabled" : true,
> "actions" : [
> {
> "name" : "compute_plan",
> "class": "solr.ComputePlanAction"
> },
> {
> "name" : "execute_plan",
> "class": "solr.ExecutePlanAction"
> }
> ]
> }
> }' http://localhost:8983/solr/admin/autoscaling
> {code}
> As soon as I add this by logs start getting spammed with these every 1 second
> {code:java}
> INFO  - 2018-06-07 15:29:16.243; [   ] org.apache.solr.servlet.HttpSolrCall; 
> [admin] webapp=null path=/admin/metrics 
> params={wt=javabin&version=2&key=solr.core.gettingstarted.shard2.replica_n4:QUERY./select.requestTimes:1minRate&key=solr.core.gettingstarted.shard1.replica_n1:QUERY./select.requestTimes:1minRate}
>  status=0 QTime=0
> INFO  - 2018-06-07 15:29:17.253; [   ] org.apache.solr.servlet.HttpSolrCall; 
> [admin] webapp=null path=/admin/metrics 
> params={wt=javabin&version=2&key=solr.core.gettingstarted.shard2.replica_n4:QUERY./select.requestTimes:1minRate&key=solr.core.gettingstarted.shard1.replica_n1:QUERY./select.requestTimes:1minRate}
>  status=0 QTime=0
> INFO  - 2018-06-07 15:29:18.263; [   ] org.apache.solr.servlet.HttpSolrCall; 
> [admin] webapp=null path=/admin/metrics 
> params={wt=javabin&version=2&key=solr.core.gettingstarted.shard2.replica_n4:QUERY./select.requestTimes:1minRate&key=solr.core.gettingstarted.shard1.replica_n1:QUERY./select.requestTimes:1minRate}
>  status=0 QTime=0
> {code}
> Does 1 second seem to aggressive ? 
> Maybe the waitFor could be taken into account and query the metrics at 
> waitFor/2 or something ? 



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

-
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org