Through CM, You'd first have to enable the Capacity Scheduler by changing the yarn.resourcemanager.scheduler.class from Fair to Capacity in YARN Configuration. Then, add the required configuration in the Capacity Scheduler Configuration (xml) file through CM in YARN Configurations. See this for more info about it. After adding the required properties, you may have to restart YARN or the ResourceManager.
... View more
Hi, Issues that create alertable events are stored in the Cloudera Management Service known as the Event Server. The Alert Publisher gets Events from the Event Server and determines whether they need to be sent out via SMTP (or SNMP, if configured) according to properties in each Event. Purging all previous Events would make it so there is no previous history from which to alert. If choosing this method though you also then lose the ability to research any of the previous Events. This may not be a problem for you and this environment. I haven't heard of anyone doing this before (for the reason you pose), but you could go into the Cloudera Management Service > Configuration > Event Server > " Maximum Number of Events in the Event Server Store" property. This controls the number of events to hold. If you set this low - say, 1000? 100? - and allow some time for the Event Server to purge all but that number, then you will have effectively deleted all previous historical events except a few. With no historical events found remaining in this index, there will be none for the cluster to alert about. Again, I've not used it in this manner but it seems like it could meet your need. Understand that this is deleting historical data, so you should determine if this is the right course of action for you. If you wish, you could also make a pre-purge copy of "Event Server Index Directory" which is in /var/lib/cloudera-scm-eventserver by default. Then if you absolutely needed to you could revert the change to the pre-purge state. Regards, -- Mark S.
... View more