Test 10'000 votes

[ineiti]

Here some bugs which still exist while testing 10'000 votes on a production system:

• make sure the votes actually work - my current script sends invalid votes
• don't lose votes, not sure which one is faster of - see also Robustness: concurrency issues dedis/d-voting#345 Robustness: votes are disappearing between the backend and the DELA node dedis/d-voting#346
  • adding a queue in the backend to only send one vote every 2 seconds
  • find the bug and fix
• check that we can do 2'000 votes
  • voting: 100ms-120ms for every ballot
  • shuffling: 5 minutes
  • decryption: 2 minutes
  • error while sending the combination: wrong nonce (2107 < 2110)
  • cycling nodes
  • combining: 2 minutes
  • results seem OK
• check that we can do 10'000 votes
  • closing: 80s
  • shuffling: failed
  • changed GetProof to GetRead
  • shuffling: 9m
  • decryption: 5m
  • combining: 1m - frontend shows timeout
  • results seem OK
• check that we can do 10'000 votes - new version w/o GetProof
  • closing: 7s
  • shuffling: times out - reloading shows that shuffling is available
  • shuffling: 11m
  • decryption: 1m - 1st failed, restarted blocks, restarted decryption
  • combining: 1m
  • results seem OK
• when getting the list of the forms in the frontend, it loads all votes, just to show the list - the slowdown was in the backend
• fixed
  • Failure when lots of data to load #126
  • check what happens if the nodes restart before/during setup
  • sometimes the viewchange fails, and I'm not sure if it works with restarting the nodes
  • sometimes one of the nodes is not available for the initialization step, and then everything fails
  • look for more Heisenbugs

To be done later:

• Shuffling and view change #130
• Frontend display of long actions #128
• Improve loading time with many blocks #127
• Viewchanges failing #129
• add progress bar during shuffling / decryption #143

[PascalinDe]

is this ticket up-to-date?

[ineiti]

Updated it and closed it. There are a couple of heisenbugs which seem to be fixable by restarting the nodes.