Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

...

  • Stop all jobs in old server, then terminate. Check that the checkpoint file was written.
  • Module load ecflow/5new
  • Start the new ecflow 5 server. It will automatically read the  checkpoint file that were created in version 4.7.X and above.

Migration( only required for ecflow < 4.7.0)

In most cases we can easily migrate from one version of ecflow to another. Since the checkpoint file will be readable by the new version of ecflow.

However occasionally this may not always be possible. (i.e. if there has been large internal re-factoring). On these occasions we can use ecflow_client --migrate.

The simplest migration involves running:

Code Block
ecflow_client --migrate > migrate.def # run on old server


migrate.def is like a normal definition file where the state is encoded as comments.
Then run the following with the new client/server:

Code Block
ecflow_client --load migrate.def # run on new server


This will load the file into the new server preserving all state information.

 If the backup servers functionality is used, then then backup servers should also be migrated at the same time.
The following notes provides more detail guidance on the migration process.
This assumes you are migrating half way through running some experiments,and want to continue where you left.

 Steps for Old server:

  • shut-down

    Code Block
    languagebash
    titleshutdown
    ecflow_client --shutdown


  • suspend all suites

    Code Block
    languagebash
    titlesuspend
    CL="ecflow_client --port 3142 --host machineX"        
    for s in $($CL --suites); do $CL --suspend /$s; done


  • wait for active/submitted tasks to complete
  • halt the server:

    Code Block
    languagebash
    titlehalt
    ecflow_client --halt


  • Use --migrate to dump state and structure to a file:

    Code Block
    languagebash
    titlemigrate
    ecflow_client --migrate > all_suites.def


  • terminate server *or* leave server running but start new server on different machine to avoid port number clash.
  • remove checkpt and backup checkpt files, to prevent new server from loading them

     *Only* applicable if starting new server on same machine

 Steps for New server:

  • module load latest release

    Code Block
    titleLoad latest ecflow
    > module load ecflow/5new
    > module load python3


  • start server
  • load the migration file:

    Code Block
    languagebash
    titleLoad
    ecflow_client --load=all_suites.def


  • set server running:

    Code Block
    languagebash
    titlerestart
     ecflow_client --restart


  • resume suspended suites:

    Code Block
    languagebash
    titleResume
    CL="ecflow_client --port 3142 --host machineX"       
    for s in $($CL --suites); do $CL --resume=/$s; done


 If you are using earlier versions, then :

...