Uploaded image for project: 'Couchbase Server'
  1. Couchbase Server
  2. MB-47591

Incorrect cgroup resource detection (core/memory) when running in Container / K8S

    XMLWordPrintable

Details

    • Triaged
    • Yes

    Description

      We need to calculate the right number of cores configured by the container environment and ensure all services use it. Since NS-SERVER is the bootstrapping all services, we'll need to add logic to detect and pass on the proper args/env as appropriate to the type of runtime a service is using.

      Most of our services uses a some flavor of managed virtual machine (Java, Golang, Erlang). These are the actual components that automatically detects number of cores, based on whichever calls they make to the OS. Some may offer a way to manually force number of cores avail to the virtual machine. Each service will need to provide bootstrapping guidance how to limit theirs. Then, we can add a wrapper to the bootstrapping that detects whether or not, cgroups is in use and calculate the cores made available to the container and pass it on to the startup script of each service and use the proper arguments to restrict the number of cores. Since most services are started by the babysitter, this can be done centrally by ns_server.

      Java : command line -XX:ActiveProcessorCount=<cores>
      Golang : environment variable GOMAXPROCS=<cores>
      Erlang: command line +S <cores>:<cores>
      C/C++: No Virtual Machine. cgroup detection logic needs to be prescribed in code.

      Important to note, as the number of threads created by each service may, or may not be derived by the number of available cores, each service will need to adjust their logic of thread allocation. In the case of Erlang based processes (NS_SERVER, COUCHDB), there will be one thread per core by default. So adjusting number of cores will also impact number of threads.

      Attachments

        Issue Links

          blocks

          Bug - A problem which impairs or prevents the functions of the product. CMOS-233 Memory statistics in containers use those of the host

          • Major - Major loss of function.
          • To Do

          Task - A task that needs to be done. K8S-2724 Add tests for cgroup/COUCHBASE_CPU_COUNT changes

          • Major - Major loss of function.
          • In Progress
          depends on

          Bug - A problem which impairs or prevents the functions of the product. MB-52029 Eventing Service should honor the CPU & Memory limits set in cgroups

          • Critical - Crashes, loss of data, severe memory leak.
          • Resolved

          Bug - A problem which impairs or prevents the functions of the product. MB-40464 CPU statistics in containers limited via CFS are misleading

          • Major - Major loss of function.
          • Closed

          Bug - A problem which impairs or prevents the functions of the product. MB-49632 cb::get_available_cpu_count ignores CPU quota

          • Major - Major loss of function.
          • Closed

          Bug - A problem which impairs or prevents the functions of the product. MB-51864 The Backup Service should honor the GOMAXPROCS env var if specified

          • Major - Major loss of function.
          • Closed

          Improvement - An improvement or enhancement to an existing feature or task. MB-51866 [cbas] Respect COUCHBASE_CPU_COUNT env (set max num of threads for JVM)

          • Major - Major loss of function.
          • Closed
          relates to

          Bug - A problem which impairs or prevents the functions of the product. MB-52686 cbbackupmgr does not correctly recognise container limits

          • Critical - Crashes, loss of data, severe memory leak.
          • Closed

          Bug - A problem which impairs or prevents the functions of the product. MB-48247 Account for container memory and cpu limits in plasma storage

          • Critical - Crashes, loss of data, severe memory leak.
          • Closed

          Bug - A problem which impairs or prevents the functions of the product. MB-49381 Projector memory monitoring doesn't work correctly in memory limited containers

          • Critical - Crashes, loss of data, severe memory leak.
          • Closed

          Bug - A problem which impairs or prevents the functions of the product. MB-40464 CPU statistics in containers limited via CFS are misleading

          • Major - Major loss of function.
          • Closed

          Task - A task that needs to be done. MB-48873 Consider runtime.GOMAXPROCS(0) vs runtime.NumCPU()

          • Critical - Crashes, loss of data, severe memory leak.
          • Closed

          Bug - A problem which impairs or prevents the functions of the product. MB-48247 Account for container memory and cpu limits in plasma storage

          • Critical - Crashes, loss of data, severe memory leak.
          • Closed

          Bug - A problem which impairs or prevents the functions of the product. MB-49382 [CBSE][sigar] Indexer CPU and memory monitoring doesn't work correctly in CPU limited containers

          • Critical - Crashes, loss of data, severe memory leak.
          • Closed

          Improvement - An improvement or enhancement to an existing feature or task. K8S-2514 [Self Cert Tool] Enhance PreFlight checks to get actual count of PID limit on containers.

          • Minor - Minor loss of function, or other problem where easy workaround is present.
          • Open
          links to

          Web Link Design Approach

          Web Link PRD

          mentioned in

          Page Loading...

          Activity

            People

              sumedh.basarkod Sumedh Basarkod (Inactive)
              meni.hillel Meni Hillel (Inactive)
              Votes:
              3 Vote for this issue
              Watchers:
              37 Start watching this issue

              Dates

                Created:
                Updated:
                Resolved:

                PagerDuty