097d82535d
* wip * commit with fewer error messages * reduce headlag from 3 to 1 * fixes for params * add composition for local:docker * revert drand-halt plan * initial monitor of miner power and chain epochs * revert params to be same as filecoin mainnet * increase timeout for drand resuming from 30sec to 120sec. * increase log level * upgrade lotus to v0.6.1 * upgrade drand to v1.1.1 * increase prepare node timeout from 1 min to 3 min * upgrade drand to master. increase timeouts for prepare drand node * nil the stmgr.ForksAtHeight map * modify starting/stopping of drand within testplan * increase drand outage to 45min. so that we miss windows * upgrade proof parameters in docker images * revert in-complete changes * use correct runtime debug image |
||
|---|---|---|
| .. | ||
| README.md | ||
| stress.god | ||
Payment channels end-to-end tests
This package contains the following test cases, each of which is described further below.
- Payment channels stress test case (
stress.go).
Payment channels stress test case (stress.go)
WIP | blocked due to https://github.com/filecoin-project/lotus/issues/2297
This test case turns all clients into payment receivers and senders. The first member to start in the group becomes the receiver. All other members become senders.
The senders will open a single payment channel to the receiver, and will
wait for the message to be posted on-chain. We are setting
build.MessageConfidence=1, in order to accelerate the test. So we'll only wait
for a single tipset confirmation once we witness the message.
Once the message is posted, we load the payment channel actor address and create
as many lanes as the lane_count test parameter dictates.
When then fetch our total balance, and start sending it on the payment channel, round-robinning across all lanes, until our balance is extinguished.
TODO:
- Assertions, metrics, etc. Actually gather statistics. Right now this is just a smoke test, and it fails.
- Implement the receiver logic.
- Model test lifetime by signalling end.