Synchronize container volume information once per epoch #105

New issue

Closed

opened 2025-12-28 17:18:26 +00:00 by sami · 0 comments

sami commented

2025-12-28 17:18:26 +00:00

Owner

Originally created by @alexvanin on GitHub (Jan 20, 2021).

Originally assigned to: @alexvanin, @cthulhu-rider on GitHub.

At start of each epoch, storage node gets collected volume estimation (#327), pick couple of rendezvous nodes within container and send them estimation.

Rendezvous nodes are picked by building placement for container based on previous epoch network map and some seed that is constant + previous epoch number.

If storage node gets such request, then it should check:

if it is rendezvous node for this container (if not, then log and ignore message),
if request sender is a node from previous epoch container (if not, then log and ignore message).
if it is already processed request (if so, ignore message),
if node finished estimation processing for this container (if so, ignore message).

Rendezvous node collects estimation per container for some time. It waits for some timeout since last correct request per container. When timeout happens, rendezvous node calculates average volume size based on 90th percentile. Then the result is saved in audit contract.

Originally created by @alexvanin on GitHub (Jan 20, 2021). Originally assigned to: @alexvanin, @cthulhu-rider on GitHub. At start of each epoch, storage node gets collected volume estimation (#327), pick couple of rendezvous nodes within container and send them estimation. Rendezvous nodes are picked by building placement for container based on previous epoch network map and some seed that is constant + previous epoch number. If storage node gets such request, then it should check: - if it is rendezvous node for this container (if not, then log and ignore message), - if request sender is a node from previous epoch container (if not, then log and ignore message). - if it is already processed request (if so, ignore message), - if node finished estimation processing for this container (if so, ignore message). Rendezvous node collects estimation per container for some time. It waits for some timeout since last correct request per container. When timeout happens, rendezvous node calculates average volume size based on 90th percentile. Then the result is saved in audit contract.