Service Availability During a Rolling Upgrade Do you wish to perform a Rolling Upgrade? : Rolling upgrades need a minimum replication factor of three (this is the default in Mammoth) to have full availability during the upgrade. You can choose a rolling upgrade or continue with a conventional upgrade. ![]() In cases where rolling upgrade is an option, Mammoth displays the following message on the console and then prompts you to chose whether or not to proceed with the installation as a rolling upgrade: The only exception is that the first time a single-rack cluster is extended onto a second rack, there will be downtime for the cluster extension (as master roles are moved to the additional rack), but subsequent cluster extensions will not require any downtime. Upgrading the Software on Oracle Big Data Applianceįor Adding Servers to a Cluster, there is normally no cluster downtime for the process. If the node has already been rebooted (and the prompt has already been answered) then the upgrade will continue without displaying the prompt again.Ī rolling upgrade is an option for the following upgrade tasks described in this document: This prompt is displayed once per node upgrade. When you are ready, resume the installation using mammoth -p. The upgrade stops so that you can fix the problems. Reboot of node mnode22bda08 had errors/warnings, do you wish to proceed anyway? : If the update detects error or warning conditions during a server reboot, then the process pauses and prompts you to continue. The reboot event will “roll” from one server to the next (or next pair) automatically if each server's post-upgrade reboot is successful. Estimate an average of 8-10 minutes per pair of non-critical nodes, in addition to 4 minutes each for the Mammoth node and critical nodes (which are rebooted one at a time).Ī rolling upgrade can be unattended. Rolling upgrades are only available for installations running CDH 4.1 or later.īecause sequential reboots of the nodes in pairs incrementally add time to the process, rolling upgrades take longer than a conventional Mammoth upgrade, in which all servers reboots are concurrent. As a result, some (though not all) services of the cluster can remain accessible without interruption during the upgrade. ![]() This allows the cluster to stay on-line throughout the entire series of reboots. A rolling upgrade reboots the nodes of a cluster in a sequence that always leaves one node out of three replicas active.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |