History

Wiki »

« Previous - Version 36/113 (diff) - Next » - Current version
Etienne Pallier, 08/11/2016 10:59 am

TODO

List of tasks by application, and general tasks (organization, tools, ...)

General tasks

  • Routine MGR:
    • Bugfix
  • Tests (intégration continue):
    • Mettre en place une exécution systématique des tests à chaque commit
      (pour faire plus simple, on peut imaginer que le serveur de test linux fasse un "git pull" toutes les 30mn, puis une exécution des tests, et l'envoi d'un email en cas de problème)
  • Installation script:
    • Si l'environnement virtuel existe déjà, demander s'il doit être recréé (par défaut "Non")
  • Pyros :
    • Doit pouvoir être démarré indifféremment AVANT (par défaut) les devices, ou APRES (actuellement, le AVANT est TODO)
    • De manière générale, doit être le plus possible "tolérant aux pannes"
    • Doit être le plus générique et donc paramétrable possible
  • Comet installation : Include in the install_requirements.sh script (by using stable official version for Python 3 (when available) on Linux (and Mac) and Windows 10
  • git : Create a "dev" branch (do not write anymore on the "master" branch)
  • Simulators :
    • on doit se rendre compte qu'ils sont "vivants" (ils seront progressivement remplacés par les vrais devices, mais resteront toujours utilisables à leur place)
      • Le Monitoring doit les interroger régulièrement sur leur statut (check_status, DONE)
      • Il doit stocker les statuts dans la BD (TODO)
      • Ces statuts doivent être affichés au fur et à mesure sur la page web du "device" correspondant (Devices, Site, Weather) (TODO)
    • quels sont les devices bloquants et non bloquants ?
      • Bloquants: si indisponibles (en panne ou pas démarrés), Pyros met en pause le sous-système d'observation
        Seules 3 fonctions restent toujours actives: alertes, monitoring, et serveur web (pour les routines et la surveillance)
        • Telescope
        • Camera VIS (???)
        • PLC (weather + observation conditions + site)
      • Non bloquants: si indisponibles (en panne ou pas démarrés), Pyros se contente d'en prendre note, et continue son fonctionnement normal tout en restant dans l'attente de leur redémarrage pour les réintégrer
        • Camera VIS (???)
        • Camera NIR
  • Users: gérer les 3 profils (admin, expert, user)
  • Doc :
    • TEST:
      • Unit tests : test of each module
      • Functional tests (with Celery and simulators): complete workflow
    • LAUNCH:
      • Start simulators
      • Start Celery
      • Start pyros
    • USE:
      • Administration of the database: http://localhost:8000/admin
      • Interacting with Pyros: http://localhost:8000
        • Watch the environment: Devices, Site, Weather
        • Watch the schedule: Schedule
        • Watch the data processing (workflow): System (Dashboard)
        • Submit a Routine Request (and get results): Routines
        • Watch alerts: Alerts
        • Simulate an Alert: Alerts (TODO: Un user "admin" doit pouvoir déclencher une alerte type depuis la page web Alerts)
        • Manual operations on the Telescope : Devices/Telescope (TODO)
        • Manage users : Users

Applications tasks

Dashboard

  • Create the backoffice views as the modules are integrated in pyros
  • Think about a system of permissions

Scheduler

  • views
    • Link the main page to the current schedule instead of the simulation page
    • Show user sequences in the schedules (with links)
    • Give accces to old schedules (day / days before / all end-night plannings / all plannings)
    • Give access to the refused sequences of a schedule, and the reasons of rejects
  • scheduler
    • Change the system to determine night start/end (they must be given in parameter, only if first_schedule is True)
    • Store the reasons of rejects (create a new attribute, in shs ?)
    • What is the 'flag' attribute ? (@AK)
    • Do not create the execute_sequence tasks if it's not the night (- 120 seconds)
    • Priority and quota computing
    • Quotas evolution
    • Blank space filling
    • At the end of a scheduling send, it to the IC ?

Alert Manager

  • Web :
    • Print if there is an alert in progress in the main page
    • Link the alerts to their status and results
  • Connect to a real VOEvent broker
  • Determine the communication with FSC for strategy change
  • VOEvents :
    • Extract the good fields (see AK Q&A 07/01/2016)
    • Fill the request & alerts objects
    • Use strategies to build a request
    • Possibility to change the default strategy
    • Handle VOEvent updates
    • Be careful to not create 2 alerts for a same GRB, seen by 2 different satellites

Analyzer

  • Apply the calibrations in the right function
  • Apply the analyses only if it's a GRB
  • Implement the analyses
  • Send analyses to FSC

Majordome

  • TaskManager
    • When a sequence is cancelled, give back the quota to the user
    • In case of alert, do not stop the ongoing plan, and make the instruments abort
  • execute_sequence
    • Add the PLC checks at start (to see if we do the slew)
    • Use the global telescope (instead of creating one here)
    • Give first_schedule as false when a scheduling is launched
    • Remove the default countdown (1, for tests)
  • system_pause
    • Abort the isntruments
    • Stop the execution tasks
  • system_restart
    • Start a scheduling
  • change_obs_conditions
    • Change sequences status (if needed)
    • If some status changed, re-launch a scheduling

Monitoring

  • views
    • Move the dashboard here
    • Print the instrument status
    • Print PLC informations (with the evolution)
    • In the dashboard screens, put scroll on each screen to see the old logs
  • Monitoring task
    • Uncomment the scheduling at the beginning
    • Implement night start/end computation
    • Initialize communication with the instruments
    • Configure intruments at start
    • Send software versions to the IC
    • Initialize connection with PLC
    • After the starting actions, loop to wait for the instruments configuration to be finished
    • Ask PLC for status
    • Ask filter wheel for status
    • Store the instruments & PLC status
    • Send all status to IC
    • Analyse PLC status (obs conditions, ...)
    • Create tasks of obs condition changes
    • At the beginning of the loop, ping the DISABLED devices. If they answer, change their status (TBD), and do their initialization (ex: If it's a camera, set its temperature to -150°)

Observation Manager

  • execute_plan :
    • Use the cameras at a global level instead of creating them here (same for the filter wheels)
    • Uncomment the instruments_ready waiting function
    • Uncomment the observation_ending waiting function
    • Try to remove code duplication
    • Determine what needs to be done at the end of an observation
  • create_calibrations :
    • Make the calibration images
    • Generate super images
    • Send them to the IC

Routine Manager

  • Web
    • Put the goods fields (for coordinates etc)
    • Only propose the objects that matches the conditions (ex: scientific programs of the user only)
    • Do all the needed checks
    • Add automatic computation of JD1/JD2
    • Add checkbox for JD / GD
    • Add options : copy my sequence on x days, and authorise report
    • Add ETC-IS simulation
    • Add help for new users (and for it the first time an account come on the page)
  • Do more checks at unserialization
  • views
    • When saving, do more checks on coordinates, jd1/2 etc
    • Uncomment filter for alerts removing
    • When submitting, use the monitoring to determine sequences status
    • When submitting, modify the first_schedule to False, when scheduling
    • When unsubmitting, uncomment the check for EXED and EXING removing
    • When unsubmitting, uncomment the scheduling and change the first_schedule to False

User Manager

  • Password recovery
  • Profile page
  • User validation by administrator / commission
  • Handle permissions and access

Common

  • Change the 'first_schedule' to False at the end of RequestBuilder.validate()

Devices

  • When a socket connection fails (exception thrown), set the device to DISABLED (or that kind of status) in the DB, and create a system_pause task (majordome)