0

I am following http://docs.mongodb.org/manual/release-notes/3.0-upgrade/ to upgrade mongodb cluster from 2.6 to 3.0. The downtime is less than 1 second? because it takes less than 1 sec to elect one secondary to become primary. Am I right? I need to double check the whole process and the downtime...

thanks.

Improve this question

1 Answer 1

Reset to default
0

If you follow the upgrade process it is designed to avoid downtime -- for example using rs.stepDown() to gracefully step down the replica set primaries rather than abruptly triggering an election by shutting down the server.

Of course there is no guarantee if something doesn't go to plan, but you should test your upgrade in a QA/staging environment before upgrading production.

Improve this answer
4
  • FYI, another option to consider in future (though probably not free depending on how large your cluster is) would be using the MongoDB Management Service (MMS)'s Automation feature. That changes all the manual steps for upgrading a cluster into a UI (or API) driven edit from your current 2.6.x version to 3.0.x: mms.mongodb.com.
    – Stennie
    Commented Mar 19, 2015 at 10:57
  • Thanks @Stennie. Now, I set up one small cluster on my own machine under ubuntu. I am not sure whether MMS can be used in my case. It seems that MMS is used with AWS. In order to monitor my cluster, is there UI tool? Do you have any other suggestions?
    – BAE
    Commented Mar 19, 2015 at 13:34
  • While MMS supports provisioning instances on AWS, you can also use it to manage an existing cluster or deploy to your own environment (i.e. AWS is not a requirement): docs.mms.mongodb.com/tutorial/nav/add-servers. There are also other options for monitoring: docs.mongodb.org/manual/administration/monitoring.
    – Stennie
    Commented Mar 19, 2015 at 19:15
  • 1
    It is to note that a stepdown leads to an election taking place. During an election, which might take up to 3 seconds (iirc), there is no primary and no data can be written. Unless the read preference is at least set to "primary preferred", reads are impossible, too. So there is at least a downtime of (number of data bearing nodes -1) * 3 seconds.
    – Markus W Mahlberg
    Commented Apr 10, 2015 at 18:13

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.