feat(server): dont use channel for replication / save df #4041
Conversation
| } | ||
|
|
||
| void RdbSaver::Impl::StartIncrementalSnapshotting(Context* cntx, EngineShard* shard, | ||
| LSN start_lsn) { | ||
| auto& db_slice = namespaces.GetDefaultNamespace().GetDbSlice(shard->shard_id()); | ||
| auto& s = GetSnapshot(shard); | ||
| s = std::make_unique<SliceSnapshot>(&db_slice, &channel_, compression_mode_); | ||
| auto on_finalize_cb = std::bind(&RdbSaver::Impl::FinalizeSnapshotWriting, this); |
There was a problem hiding this comment.
it's possible to do like this. not critical but when you pass two lambdas, maybe it's a sign that using an virtual interface is a better pattern here, and make RdbSaver::Impl derive from it.
| void SaveStagesController::SaveCb(unsigned index) { | ||
| if (auto& snapshot = snapshots_[index].first; snapshot && snapshot->HasStarted()) | ||
| void SaveStagesController::SaveBody(unsigned index) { | ||
| CHECK(!use_dfs_format_ || index == shard_set->size()); |
There was a problem hiding this comment.
this CHECK is weird index == shard_set->size()
There was a problem hiding this comment.
we use SaveBody for rdb save and for summary file in df save.
We can do several refactoring here for more readable code but I want this pr to be not so big
There was a problem hiding this comment.
added a comment for now
src/server/snapshot.h
Outdated
| @@ -85,7 +76,7 @@ class SliceSnapshot { | |||
|
|
|||
| // Waits for a regular, non journal snapshot to finish. | |||
| // Called only for non-replication, backups usecases. | |||
| void Join() { | |||
| void WaitSnapshoting() { | |||
There was a problem hiding this comment.
WaitSnapshoting -> WaitSnapshotting
please check if you have more shoting cases
I will do addittional refactoring changes in a separate PR ok? |
|
Yeah sure, it was a general comment 👌 |
Signed-off-by: adi_holden <[email protected]>
Signed-off-by: adi_holden <[email protected]>
Signed-off-by: adi_holden <[email protected]>
Co-authored-by: Roman Gershman <[email protected]> Signed-off-by: adiholden <[email protected]>
Co-authored-by: Roman Gershman <[email protected]> Signed-off-by: adiholden <[email protected]>
Co-authored-by: Roman Gershman <[email protected]> Signed-off-by: adiholden <[email protected]>
Signed-off-by: adi_holden <[email protected]>
adiholden
force-pushed
the
snapshot_write_to_sink
branch
from
549ea37 to
5c14022
Compare
* feat server: dont use channel for replication / save df Signed-off-by: adi_holden <[email protected]>
fixes #4040