[ https://issues.apache.org/jira/browse/CASSANDRA-14395?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16443366#comment-16443366 ]
Kurt Greaves commented on CASSANDRA-14395: ------------------------------------------ I very much like the idea, and think the plan so far makes sense to limit it to a few basic operations/checks. Questions that come to mind: # Seeing as this is a sidecar, are we intending to run it on all nodes? To me a management process only needs to run on a single node; potentially a subset for HA. Health checks make sense to be running on every node, but a process to issue bulk nodetool commands to the entire cluster not so much (similar for repair). # How are we going to orchestrate bulk management commands? Seems like JMX is the obvious choice, but what are we using to distribute commands across the cluster? Having every node interconnected on JMX seems like a really bad idea, so do we pass commands between the management processes and then each node issues its own JMX commands to itself? ## How do we handle conflicting commands from separate nodes? ## What's the overhead of running this process on every node when it only really needs to be run on one? I think need a bit more detail in the proposal to get a better feel of what this looks like. From my perspective ATM it's starting to look like 2 separate sidecars would make more sense, one for every node (health checks, compactions in the future, backup, etc) and then one for cluster-wide management (bulk commands, restore, repair in the future) but this could just be because I've missed the point. > C* Management process > --------------------- > > Key: CASSANDRA-14395 > URL: https://issues.apache.org/jira/browse/CASSANDRA-14395 > Project: Cassandra > Issue Type: New Feature > Reporter: Dinesh Joshi > Assignee: Dinesh Joshi > Priority: Major > > I would like to propose amending Cassandra's architecture to include a > management process. The detailed description is here: > https://docs.google.com/document/d/1UV9pE81NaIUF3g4L1wxq09nT11AkSQcMijgLFwGsY3s/edit > I'd like to propose seeding this with a few simple use-cases such as Health > Checks, Bulk Commands with a simple REST API interface. -- This message was sent by Atlassian JIRA (v7.6.3#76005) --------------------------------------------------------------------- To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org For additional commands, e-mail: commits-h...@cassandra.apache.org