You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@helix.apache.org by "Zhen Zhang (JIRA)" <ji...@apache.org> on 2014/05/14 02:13:14 UTC
[jira] [Updated] (HELIX-444) add per-participant partition count
gauges to helix
[ https://issues.apache.org/jira/browse/HELIX-444?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Zhen Zhang updated HELIX-444:
-----------------------------
Description:
We need a way to pull the known down partition counts out of DifferenceWithIdealState when an instance is offline, reducing the alert volume to solely the down instance notification. Without metrics from helix indicating the number of partitions hosted on a given participant, we can't reason as to which "DifferenceWithIdealState" counts are supposed to be down and which are an actually difference caused by something other than a node outage.
These should be produced on a per-participant, per-resource basis (ie., helix.i001.participantstatus.cluster.host.db.partitiongauge = 64 or whatever)
was:
We need a way to pull the known down partition counts out of DifferenceWithIdealState when an instance is offline, reducing the alert volume to solely the down instance notification. Without metrics from helix indicating the number of partitions hosted on a given participant, we can't reason as to which "DifferenceWithIdealState" counts are supposed to be down and which are an actually difference caused by something other than a node outage.
These should be produced on a per-participant, per-resource basis (ie., helix.i001.participantstatus.ESPRESSO_USCP.ela4-app1234.prod.linkedin.com_11932.ucpx.partitiongauge = 64 or whatever)
> add per-participant partition count gauges to helix
> ---------------------------------------------------
>
> Key: HELIX-444
> URL: https://issues.apache.org/jira/browse/HELIX-444
> Project: Apache Helix
> Issue Type: Improvement
> Reporter: Zhen Zhang
> Assignee: Zhen Zhang
>
> We need a way to pull the known down partition counts out of DifferenceWithIdealState when an instance is offline, reducing the alert volume to solely the down instance notification. Without metrics from helix indicating the number of partitions hosted on a given participant, we can't reason as to which "DifferenceWithIdealState" counts are supposed to be down and which are an actually difference caused by something other than a node outage.
> These should be produced on a per-participant, per-resource basis (ie., helix.i001.participantstatus.cluster.host.db.partitiongauge = 64 or whatever)
--
This message was sent by Atlassian JIRA
(v6.2#6252)