large_scale_long_lived and large_scale_model genny workloads potentially faulty

XMLWordPrintableJSON

    • Type: Improvement
    • Resolution: Unresolved
    • Priority: Major - P3
    • None
    • Affects Version/s: None
    • Component/s: None
    • None
    • None
    • None

      While working on PM-1986 alex.cameron wrote a new workload based on large_scale_long_lived and large_scale_model workloads. I was testing if the workload distributes reads/writes binomially across the _id keyspace and found reads and writes actually do not end up going to WiredTiger.

      The filters we use on _id for reads and updates assume _id to be numeric, whereas the load phase lets _id to be defaulted to it's ObjectID representation. We created WT-7041 to fix this bug in the new workload.

      Having based new workload on large_scale_long_lived and large_scale_model, it is very likely that these workloads are not doing the right thing at all, and doing reads/writes with filters that do not match any documents.

      This ticket is to check if the workloads have been faulty and if yes, fix them. WT-7041 could hint at the possible solution.

            Assignee:
            [DO NOT USE] Backlog - Storage Engines Team
            Reporter:
            Sulabh Mahajan
            Votes:
            0 Vote for this issue
            Watchers:
            8 Start watching this issue

              Created:
              Updated: