ExportXMLWordPrintableJSON

    • Type: Improvement
    • Resolution: Fixed
    • Priority: Major - P3
    • 9.1.0-rc1026
    • Affects Version/s: None
    • Component/s: None
    • None
    • Storage Engines - Server Integration
    • Fully Compatible
    • SESIroflcopter 2026-11-03, SESIWhinery 2026-11-17
    • None
    • None
    • None
    • None
    • None
    • None
    • None

      Calls from mongod to CMS currently use a heartbeat to detect dead servers and network errors. There is still a possible gap where a CMS connection is accepted and established but takes a long time to service the request (the connection gets stuck and theres no response) In that case calls like GetLogSegmentList can leave startup waiting on the stuck CMS instead of trying another cell. If we had something like a 30s timeout for each attempt we could fallback

      The ORP change in SERVER-135177 gave each attempt a 30s timeout (kResolveORPEndpointRpcAttemptTimeout). This shouldn't live specially in the ORP code though we should make it live in the shared CMS code so we 1. don't override the retryPolicy() and 2. for consistency within CMS calls.

      We should move the per attempt timeout into CMSChannelManager and use it for both ORP resolving endpoint and GetLogSegmentList. When an attempt times out we can treat the server as unavailable and try the next one within the total wait deadline from the caller. Keep existing behaviour for other timeout errors

            Assignee:
            Nic Hollingum
            Reporter:
            Clarisse Cheah
            Votes:
            0 Vote for this issue
            Watchers:
            2 Start watching this issue

              Created:
              Updated:
              Resolved: