Issues Identified in MMCloud Half Moon Bay 2.5 Release

Known Issues and Limitations

  • Summary statistics reported by OpCenter and reported by the Seqera platform may differ

    A Nextflow pipeline may include multiple executors (including the local host). The OpCenter can only track the tasks where float is the executor. The summary statistics shown when the pipeline completes (number of successes, number of failures, and so on) may be reported differently by the OpCenter and by the Seqera platform. For detailed tracking of all tasks in a pipeline, see the Nextflow log for that pipeline (.nextflow.log).

  • No matches found if "=" operator used to filter on fields where the number of significant digits is greater than shown in the display (CLI or web interface). Use ">=" or "<=" instead.
    Example: 
    float list -f "submitTime=2024-02-05T15:02:05Z"
No jobs
float list -f "submitTime>=2024-02-05T15:02:05Z"
+-----------------------+-...--+----------------------+------------+
|          ID           | ... |     SUBMIT TIME      |    COST    |
+-----------------------+-...--+----------------------+------------+
| mmmi69sbfax9wdlr0ajop | ... | 2024-02-05T15:02:05Z | 3.1711 USD |
+-----------------------+-...--+----------------------+------------+
  • Insufficient local disk cache causes errors when accessing s3fs-mounted volumes

    This is a known issue with s3fs. The workaround is to configure a large root volume on the worker node. Use --rootVolSize 80g with float submit to double the root volume size from the default of 40G.

  • S3 buckets with a "." in the name (for example, somename.memverge) cannot be used as data volumes

    Although AWS permits the use of "." in S3 bucket names, jobs submitted with a data volume defined by an S3 bucket with a name containing a "." (for example, --dataVolume [mode=rw]s3://somename.memverge:/data) fail to execute because the data volume cannot be mounted.

  • A failed periodic snapshot may not clean up the associated file system snapshot

    A periodic snapshot captures a snapshot of the memory state and a file system snapshot, and stores both on a dedicated storage volume. If the in-memory snapshot fails and the file system snapshot succeeds, the file system snapshot may persist on the storage volume even though it is not useful by itself. Delete the file manually.

  • TCP connections are not restored when a job is recovered

    A job with active external TCP connections may fail to resume when paused and recovered on a new VM. To ensure the job recovers, submit the job with the --extraOptions “--tcp-close“ option included. This ensures that the job recovers although the TCP connections are not restored. If the application running inside the container requires the TCP connections, the application establishes new connections.

  • No support for multiple squashfs file systems

    A loopback device portrays a file as a block device so that it can be mounted as a file system. Each squashfs file system requires a loopback device. Currently, only one loopback device can be created in the container.

  • Using the CLI command float config ldap --anonymous=true to configure anonymous binding for LDAP fails if the LDAP configuration has been reset. Use the float config ldap --conf conf.yaml method instead.