Skip to content

Conversation

@martinsumner
Copy link
Contributor

@martinsumner martinsumner commented Feb 2, 2025

See #8

This changes the riak_core_vnode update_never_started so that it prefers to start vnodes that are not owned by this node within the ring. These vnodes will start in series - so this makes it more likely that vnodes owned (and containing data) will start in parallel.

Changes logging so that we log the starting of the vnode, not of the vnode worker pool.

Further logging of the handoff completion process has been added. This is as a result of OpenRiak/riak_kv#23 - an issue which is not currently resolvable due to a lack of log information about what is occurring at handoff completion

Make the default number of vnode to start per management tick higher.  Still start them in series, but prefer those not owned in the current ring - and so less likely to contain data.

This makes it more likely that the `maybe_ring_changed/4` function will start the active vnodes in parallel as expected.
Do we need to know about vnode pools starting - but not vnodes?
On handoff completion log the reason why a completion is being ignored (i.e. no ring update is being made).  Also log that a ring update has been triggered.

Also with the leveled backend quieten logs associated with monitoring of the leveled backend when unregistering.

The deletion of a backend may possibly be time-consuming when a system is under stress, and the vnode will be blocked for this time - so log the time it takes to delete the backend to help with monitoring.
Co-authored-by: Thomas Arts <thomas.arts@quviq.com>
@martinsumner martinsumner merged commit c803aea into openriak-3.4 Mar 20, 2025
2 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants