Skip to content

Stuck handoff after disabling search #455

Closed
engelsanchez opened this Issue Nov 20, 2013 · 3 comments

3 participants

@engelsanchez

It was reported that Riak ended up with handoff transfers stuck in the awaiting state through this sequence of actions:

  • The cluster had search enabled, although search wasn't actively used
  • Search merge index corruption was stalling handoffs
  • Nodes were restarted to disable search, search data was removed
  • Once back up, transfers that were waiting for search handoff to finish now are stuck in that state.

Maybe we need to do a bit more to detect invalid handoff state after situations like this.

@jrwest
jrwest commented Mar 24, 2014

what do we want to do with this issue given search is deprecated. I'm inclined to close it. @rzezeski?

@rzezeski

If we continue work on the 1.4.x series then this may be something we want to address. I personally would vote to take the lazy approach and wait until it becomes an issue for someone again.

@jrwest
jrwest commented Mar 24, 2014

I'm going to go ahead and close this out. 1.4.x search is unlikely to see improvements, unless it affects users directly, and 2.0 search does not work in the same manner and does not have its own vnode, which is what would be stalling the handoff. Other riak_core applications could perhaps benefit (riak_pipe not really being one of them) but I don't see this as planned to be fixed. We can re-open if it comes back up.

@jrwest jrwest closed this Mar 24, 2014
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Something went wrong with that request. Please try again.