Investigate sharded DRM document loss when $limit 1 meta-cursor disposal kills an in-flight getMore during Exchange load

XMLWordPrintableJSON

    • Type: Bug
    • Resolution: Fixed
    • Priority: Major - P3
    • 9.0.0-rc0
    • Affects Version/s: None
    • Component/s: None
    • None
    • Query Integration
    • Fully Compatible
    • ALL
    • None
    • None
    • None
    • None
    • None
    • None
    • None

      On sharded clusters, a $$SEARCH_META-binding aggregation silently drops one shard's entire document stream when results span multiple getMore batches. The $limit 1 stage in the meta pipeline proactively disposes its cursors after the first meta doc, which cancels the other shard's in-flight meta getMore mid-Exchange::loadNextBatch, setting an error on the shared Exchange. The doc cursor's subsequent getMore then fails with ExchangePassthrough, which the AsyncResultsMerger silently treats as clean EOF, returning a partial result with no error.

            Assignee:
            Daniel Segel
            Reporter:
            Daniel Segel
            Votes:
            0 Vote for this issue
            Watchers:
            2 Start watching this issue

              Created:
              Updated:
              Resolved: