From d765ff5e75245d4628efc0d85b88fbd50ef6ccfe Mon Sep 17 00:00:00 2001 From: fjy Date: Thu, 17 Jul 2014 14:53:26 -0700 Subject: [PATCH] docs for updating --- docs/content/Rolling-Updates.md | 44 +++++++++++++++++++++++++++++++++ docs/content/toc.textile | 1 + 2 files changed, 45 insertions(+) create mode 100644 docs/content/Rolling-Updates.md diff --git a/docs/content/Rolling-Updates.md b/docs/content/Rolling-Updates.md new file mode 100644 index 00000000000..a6d5cd005bd --- /dev/null +++ b/docs/content/Rolling-Updates.md @@ -0,0 +1,44 @@ +--- +layout: doc_page +--- + + +Rolling Updates +=============== + +For rolling Druid cluster updates with no downtime, we recommend updating Druid nodes in the following order: + +1. Historical Nodes +2. Indexing Service/Real-time Nodes +3. Broker Nodes +4. Coordinator Nodes + +## Historical Nodes + +Historical nodes can be updated one at a time. Each historical node has a startup time to memory map all the segments it was serving before the update. The startup time typically takes a few seconds to a few minutes, depending on the hardware of the node. As long as each historical node is updated with a sufficient delay (greater than the time required to start a single node), you can rolling update the entire historical cluster. + +## Standalone Real-time nodes + +Standalone real-time nodes can be updated one at a time in a rolling fashion. + +## Indexing Service + +### With Autoscaling + +Overlord nodes will try to launch new middle manager nodes and terminate old ones without dropping data. This process is based on the configuration `druid.indexer.runner.minWorkerVersion=#{VERSION}`. Each time you update your overlord node, the `VERSION` value should be increased. + +### Without Autoscaling + +Middle managers can be updated in a rolling fashion based on API. + +To prepare a middle manager for update, send a POST request to `/druid/worker/v1/disable`. The overlord will now no longer send tasks to this middle manager. + +Current tasks will still try to complete. To view all existing tasks, send a GET request to `/druid/worker/v1/tasks`. When this list is empty, the middle manager can be updated. After the middle manager is updated, it is automatically enabled again. You can also manually enable middle managers POSTing to `/druid/worker/v1/enable`. + +## Broker Nodes + +Broker nodes can be updated one at a time in a rolling fashion. There needs to be some delay between updating each node as brokers must load the entire state of the cluster before they return valid results. + +## Coordinator Nodes + +Coordinator nodes can be updated in a rolling fashion. \ No newline at end of file diff --git a/docs/content/toc.textile b/docs/content/toc.textile index a4769a6b2f5..76ba3e31133 100644 --- a/docs/content/toc.textile +++ b/docs/content/toc.textile @@ -17,6 +17,7 @@ h2. Getting Started h2. Booting a Druid Cluster * "Simple Cluster Configuration":Simple-Cluster-Configuration.html * "Production Cluster Configuration":Production-Cluster-Configuration.html +* "Rolling Cluster Updates":Rolling-Updates.html h2. Configuration * "Common Configuration":Configuration.html