asagi_archive_auto_failover_notes.txt

asagi_archive_auto_failover_notes.txt
Notes on auto-failover script.




===== PLAN =====

Periodically test if the archive is working.
If it is not working, trigger some arbitary code to trigger failover.
Failover refers to automatic reconfiguration of the host from its main job to active archival.




+++ Questions about how to do it +++
How do we differentiate between 4chan breaking and the archive breaking?
    Check 4chan's API for updates?
    Do we need to store more information than the highest post_num and image_num?


What triggers should be used to detect failure?
    Lack of new posts?
    Highest seen post_num not increasing in a given time period?
    Error pages/codes accessing archive.
    Exceptions, use pokemon-style exception handling to trigger failover on bugs.



What are sensible thresholds?
    Do these change based on the board being observed?
    Use most active board in the archive as the failure-canary?
    Should more than one board be tested? How many boards need to fail before triggering failover code?



What sort of response should occur on failure?
    Leave this to the admin running the script to decide, simply provide ability to run arbitary shell commands.





===== Basic high-level pseudocode =====



# Load archive API JSON











===== USAGE =====

TODO