Obey "throttle" by checking server health state

[tobiasKaminsky]

Latest idea:

• clients define "bulk/high server load" operation
• before they do them (e.g. upload 100 items), they check new endpoint
• e.g. /health -> 200 or 509(?)
• only continues if "200"

Server:

• another monitoring tool monitors NC load
• if load is too high, a response header is set: X-NC-THROTTLE
• How to do
  • NC sets a header to each response
  • Apache/Nginx can do that
• client should react on it
  • once received, throttle down or show correct info to user (e.g. when happening on manual upload)
  • if next request is fine again, unthrottle
• Mobile
  • not on user initiated actions (e.g. browsing, uploading)
  • only on background jobs
    • internal two way sync
    • auto upload
• Desktop
  • to discuss

[tobiasKaminsky]

Another idea: have a response header, to not kill reuquest with destructive 429

[tobiasKaminsky]

Important: unthrottle also at random time (easy approach: have a random timespan for waiting time)

[tobiasKaminsky]

Question:
Can this work?

• Monitoring tool sees trouble
• tool issues "occ notifyClientThrottle on"
• all responses now have a X-NC-THROTTLE header <-- is this easy/with less time effort to do?

[alperozturk96]

The following operations can be resource-intensive, as they may involve bulk processing or repeated execution:

• Fetching metadata via the metadata worker
• Removing multiple files
• Uploading multiple files sequentially
• Clearing the trash bin in a single operation
• Moving or copying multiple files
• Downloading entire directories
• Thumbnail generation tasks in the Media tab
• Fetching shares information