feat(op-batcher): altda->ethda failover #13
Conversation
samlaf
force-pushed
the
samlaf/feat--op-batcher-altda-failover-to-ethda
branch
2 times, most recently
from
adfa7ce to
61d7578
Compare
samlaf
changed the base branch from
develop
to
feat--multiframe-altda-channel
|
At the high level, it usually takes us about 1 hour to fix the problem |
| @@ -130,6 +134,10 @@ func (s *FakeDAServer) HandleGet(w http.ResponseWriter, r *http.Request) { | |||
|
|
|||
| func (s *FakeDAServer) HandlePut(w http.ResponseWriter, r *http.Request) { | |||
| time.Sleep(s.putRequestLatency) | |||
| if s.failoverCount > 0 { | |||
There was a problem hiding this comment.
what is the point to decrement failoverCount, then actually handle the put
There was a problem hiding this comment.
Is it just to simplify testing?
There was a problem hiding this comment.
Because we only want to cause a temporary failure. So this way we can say "pretend you are down for the next failoverCount calls", and after that the system becomes normal again.
op-e2e/system/altda/failover_test.go
Outdated
| ) | ||
|
|
||
| // TestBatcher_FailoverToEthDA_FallbackToAltDA tests that the batcher will failover to ethDA | ||
| // if the da-server returns 503, and then fallback to altDA once altDA is available again |
There was a problem hiding this comment.
we always try altda first for every dispersal, then retry for non-altDA after sufficient retry. Wording seems odd.
There was a problem hiding this comment.
Changed the wording: 6928901
|
|
||
| countEthDACommitment := uint64(0) | ||
|
|
||
| // Most likely, sequence of blocks will be: altDA, ethDA, ethDA, altDA, altDA, altDA. |
There was a problem hiding this comment.
why most likely? And why two ethDA in a row, because we set failoverCount=2, but why altDA in the beginning?
There was a problem hiding this comment.
Check this new comment: 71d8697
samlaf
force-pushed
the
feat--multiframe-altda-channel
branch
from
1a84f49 to
bb0ba93
Compare
in TestBatcher_FailoverToEthDA_FallbackToAltDA
samlaf
force-pushed
the
samlaf/feat--op-batcher-altda-failover-to-ethda
branch
from
71d8697 to
a881df5
Compare
DON'T MERGE!
Plan is to upstream this to op's repo. Created here to get review from the team first.
This PR builds on top of the feat--multiframe-altda-channel changes (already upstreamed to op's repo, waiting for review there)
It contains 2 commits:
Right now failover is done to calldata txs because that was trivial whereas failing over to blobs or their auto mode that switches between blobs and calldata would need a nontrivial refactor and some thinking. Not sure its worth putting effort into this atm given that the whole point of failover is that it should happen very rarely and also not last very long, but let me know if you guys think otherwise.