• DougPiranha42@lemmy.world
    link
    fedilink
    English
    arrow-up
    58
    arrow-down
    21
    ·
    1 day ago

    Interesting. A client side bug causes server-level overload but only after unrelated changes gradually make the problem worse. Rerouting traffic to healthy servers overloads those too and spreads the outage. The client code surely “worked on my computer” lol

    • gian @lemmy.grys.it
      link
      fedilink
      English
      arrow-up
      1
      ·
      7 hours ago

      Probably.

      Joking aside, the problem with testing is that you normally cannot have a complete copy of the production env to test against, with all the possible cases.
      You may have a test env that do not trigger certain corner case scenario which can be dependent on something completely out of the domain, like in this case.

      If they learned something, they would add some way of testing even this kind of scenario in the future.

      • threeonefour@piefed.ca
        link
        fedilink
        English
        arrow-up
        50
        arrow-down
        2
        ·
        1 day ago

        It says multiple times that the issue was with Microsoft’s internal cloud server software, not Windows.

        The defect came from a fix prepared specifically for Microsoft 365 networking gateway servers.

        The failure was within Microsoft’s cloud infrastructure, not customer Windows systems

        • eleijeep@piefed.social
          link
          fedilink
          English
          arrow-up
          9
          ·
          edit-2
          20 hours ago

          Yeah exactly, I don’t understand how the original comment is still getting upvotes.

          Since your comment it gained (+36, -11). People don’t read, I guess.

      • DougPiranha42@lemmy.world
        link
        fedilink
        English
        arrow-up
        2
        ·
        14 hours ago

        I guess I misunderstood it.

        Inside an internal Windows process, competing tasks were waiting for access to shared resources, a problem called lock contention. Under certain traffic conditions, this exhausted processor and memory capacity.

        I read this as the behavior that precipitated the issue on the gateway was caused by how Windows sent traffic to exchange servers. So by Windows they mean whatever is running on the gateway?