Hi Angel, Thanks for your kind words.
> Is true that It is not possible to use Ambari with an existing Hadoop > installation? So, it is necessary to install all the platform using Ambari. This could be a problem for our platform team. That is true. Currently, you cannot directly manage and monitor an existing Hadoop cluster that was installed outside of Ambari. However, some folks have successfully performed "take over" procedures to bring such existing clusters under Ambari's management. This involves a bit of surgery such as using Ambari's Install Wizard on the existing cluster but using fake directories for DataNodes (and others), and switching the directories to the actual ones after Install Wizard is done, etc. This is not very straightforward and can get you in trouble if you are not careful, but it is an option. Yusaku From: Angel Cervera Claudio <[email protected]<mailto:[email protected]>> Reply-To: "[email protected]<mailto:[email protected]>" <[email protected]<mailto:[email protected]>> Date: Monday, March 23, 2015 5:37 AM To: "[email protected]<mailto:[email protected]>" <[email protected]<mailto:[email protected]>> Subject: Re: Couple of question. Hi Yusaku. Awesome community support! With people like you, I am sure that Ambari will be the best monitor tool! My last question (I don't want bored you :) ) : Is true that It is not possible to use Ambari with an existing Hadoop installation? So, it is necessary to install all the platform using Ambari. This could be a problem for our platform team. Regards. 2015-03-21 1:01 GMT+00:00 Yusaku Sako <[email protected]<mailto:[email protected]>>: Hi Angel, > Ganglia is going to be replace by Ambari Metric System. Will be the same for > Nagios? Yes, Ambari will ship with its own alerting system in 2.0. Nagios is no longer supported. > If we can not use Ambari, we are thinking in develop an adhoc application for > metering and monitoring. We are thinking in use Time Serie Databases to store > metric, like Druid or OpenTSDB. Is Amabari going to use any databases of this > type? Ambari's Metric System (AMS) uses HBase as its implementation, like OpenTSDB. In addition, AMS uses Phoenix to support SQL queries. I believe the storage layer for AMS was designed so that it can be swapped out, but I will let others who are more familiar comment on that. > And the question that maybe you are tired to listen. :) Do you know when is > planed to release the version 2.0.0? A release vote for 2.0 just went out. If everything goes well, 2.0 will ship within days. Thanks, Yusaku From: Angel Cervera Claudio <[email protected]<mailto:[email protected]>> Reply-To: "[email protected]<mailto:[email protected]>" <[email protected]<mailto:[email protected]>> Date: Friday, March 20, 2015 6:56 AM To: "[email protected]<mailto:[email protected]>" <[email protected]<mailto:[email protected]>> Subject: Re: Couple of question. Hi Yusaku. Thank you for your response. Other three question: * Ganglia is going to be replace by Ambari Metric System. Will be the same for Nagios? * If we can not use Ambari, we are thinking in develop an adhoc application for metering and monitoring. We are thinking in use Time Serie Databases to store metric, like Druid or OpenTSDB. Is Amabari going to use any databases of this type? * And the question that maybe you are tired to listen. :) Do you know when is planed to release the version 2.0.0? Regards and thanks for your time. 2015-03-19 6:42 GMT+00:00 Yusaku Sako <[email protected]<mailto:[email protected]>>: Hi Angel, On Ambari 1.7.0 and earlier, Ganglia is used to capture information about memory usage, network usage, and other system metrics, as well as Hadoop-specific metrics. Ambari 2.0.0 and onwards, Ambari Metric System will replace Ganglia as the underlying metrics collection framework. In either system, you can tune sampling/aggregation frequencies and retention policies so that you can control how long the collected samples/aggregated values are kept at what granularity. Yusaku From: Angel Cervera Claudio <[email protected]<mailto:[email protected]>> Reply-To: "[email protected]<mailto:[email protected]>" <[email protected]<mailto:[email protected]>> Date: Thursday, March 19, 2015 4:10 AM To: "[email protected]<mailto:[email protected]>" <[email protected]<mailto:[email protected]>> Subject: Re: Couple of question. Hi Yasuku. Thanks for your responses. Respect Historical data: One of the requirements that we have is to be able to show/compare historical information. For example, it is necessary to be able to compare the memory used yesterday (when we executed the new version of a heavy MapReduce process) with the memory consumed two month ago. This is only one example. Regards. 2015-03-18 6:35 GMT+00:00 Yusaku Sako <[email protected]<mailto:[email protected]>>: Hi Angel, Please see my response inline below: From: Angel Cervera Claudio <[email protected]<mailto:[email protected]>> Reply-To: "[email protected]<mailto:[email protected]>" <[email protected]<mailto:[email protected]>> Date: Friday, March 6, 2015 12:50 AM To: "[email protected]<mailto:[email protected]>" <[email protected]<mailto:[email protected]>> Subject: Couple of question. Hi everybody. I am looking for the way of custom Ambari with new widgets and views. Researching for few days (I never worked before with this tool), I have a couple of questions: * Is there any easy way to create new widget to add in the main page of Ambari? * Is there any possibility to have a different layout (different widgets and views) per user. Something like profiles. * Is there possible to secure by roles and widgets/views/services (authorization)? [YS] These will be covered by new features being worked on to make the dashboards customizable via https://issues.apache.org/jira/browse/AMBARI-9792. You will be able to define a default layout, as well as customized layouts per user, create new widgets based on different metrics, share those widgets, etc. Views (and their features) can be authorized per the Ambari Views framework already. * I did not found anything about history data (real historical data, per years). Is it possible? [YS] Can you clarify what you mean? Do you mean various time-series data emitted by the system/various components? I downloaded the source code for 1.7 (github tag 1.7), but in the pom.xml said that it is the version 1.3.0-SNAPSHOT. Curiosly, there are not release 1.3 mention in the ambari.apache.org<http://ambari.apache.org> page. Why? [YS] That's just an artifact of the source code in Ambari 1.7.0. If you want to rebuild, you can issue "mvn versions:set -DnewVersion=1.7.0.0" to set the version in the pom.xml files. You are correct that there never was a 1.3 release. Regards.
