Open olegrok opened 5 days ago
Keep in mind that in the same space can be multiple buckets stored. A possible workaround is a temporary wrap of user triggers into internal triggers which would invoke the user triggers only on changes coming externally.
Such triggers can prevent writing to spaces or handle such writes in a quite weird way. So seems vshard should disable triggers (before_replace/on_replace) that user sets to avoid such confusing behaviour.
Rebalancing is quite rare action and it can suddenly shoot even on production. For now vshard rebalancing should be explicitly tested to prevent such cases. It's not good from UX point because newcomers should deeply understand how vshard internals work.