Split roundstate saving in two #2046
Conversation
|
Coverage from tests in coverage: 51.0% of statements across all listed packagescoverage: 63.2% of statements in consensus/istanbul coverage: 42.7% of statements in consensus/istanbul/announce coverage: 56.0% of statements in consensus/istanbul/backend coverage: 0.0% of statements in consensus/istanbul/backend/backendtest coverage: 24.3% of statements in consensus/istanbul/backend/internal/replica coverage: 66.0% of statements in consensus/istanbul/core coverage: 50.0% of statements in consensus/istanbul/db coverage: 0.0% of statements in consensus/istanbul/proxy coverage: 64.4% of statements in consensus/istanbul/uptime coverage: 51.8% of statements in consensus/istanbul/validator coverage: 79.2% of statements in consensus/istanbul/validator/random |
|
|
| @@ -40,6 +41,7 @@ const ( | |||
| dbVersionKey = "version" // Version of the database to flush if changes | |||
| lastViewKey = "lastView" // Last View that we know of | |||
| rsKey = "rs" // Database Key Pefix for RoundState | |||
| rcvdKey = "rcvd" // Database Key Prefix for rcvd messages from the RoundState (split saving) | |||
There was a problem hiding this comment.
why not have differente keys for prepares, commits and parent commits?
There was a problem hiding this comment.
I believe they shouldn't have an outstanding size, though I agree that further splitting is better. I'll check how the code looks with it.
hbandura
force-pushed
the
hbandura/split_rs_db
branch
from
0dafa70
to
025eb88
Compare
Codecov ReportPatch coverage:
Additional details and impacted files@@ Coverage Diff @@
## master #2046 +/- ##
==========================================
+ Coverage 54.20% 54.22% +0.01%
==========================================
Files 692 692
Lines 115463 115618 +155
==========================================
+ Hits 62585 62691 +106
- Misses 49050 49082 +32
- Partials 3828 3845 +17
... and 25 files with indirect coverage changes Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here. ☔ View full report in Codecov by Sentry. |
| defer ms.messagesMu.Unlock() | ||
| var added = 0 | ||
| for _, msg := range msgs { | ||
| err := ms.checkValidToAdd(msg) |
There was a problem hiding this comment.
Should we log the invalid messages or the possible overrides?
There was a problem hiding this comment.
I don't think so, since it's actually expected that there will be some overrides on a load
| view := newView(2, 1) | ||
| rs := newRoundState(view, valSet, valSet.GetByIndex(0)) | ||
| rs.AddPrepare(mockViewMsg(view, istanbul.MsgPrepare, valSet.GetByIndex(0).Address())) | ||
| rs.AddCommit(mockViewMsg(view, istanbul.MsgPrepare, valSet.GetByIndex(0).Address())) |
There was a problem hiding this comment.
Shouldn't this be a istanbul.MsgCommit? (same with the next one)
hbandura
force-pushed
the
hbandura/split_rs_db
branch
from
5b297b3
to
bb16fb8
Compare
* Split roundstate saving in two: prepares/commit/parentCommit messages on one side.
* Split roundstate saving in two: prepares/commit/parentCommit messages on one side.
Split the roundstate saving, allowing consensus message processing to be a bit faster.