Jan

@riversand963 : can you already estimate if the changes to TransactionLogIterator can be made in the near/mid-term future, and when? This would greatly help because this issue keeps us from...

@riversand963 : ok, any progress on this issue is highly appreciated from our side! Thanks!

@riversand963 : wondering if a fix for this can be made for one of the upcoming releases 7.6 or 7.7. It would be super helpful because it would enable us...

@riversand963 : any idea if/when this issue can be investigated or fixed? Thanks!

This issue is still blocking the adoption of SstFileWriter-enabled code, which would unlock large performance benefits for us if using the SstFileWriter and GetUpdatesSince were made compatible. Thanks a lot...

@hemin1020 : regarding this custom Comparator code you shared... ``` int Compare(const rocksdb::Slice& a, const rocksdb::Slice& b) const override { // "extentX" < "streamX" if(a[0] > b [0]) { return...

The PR will change `DBImpl::FlushMemTable()` to return status TryAgain early if writes have stopped, but _only_ if `FlushOptions.wait` is `false`. I think that even after this PR is merged, some...

@riversand963 : the problem is that the `FlushOptions` instance used here is created internally by RocksDB, so there is no way for the caller/embedder to set `wait = false` on...

> @jsteemann I think if `CompactRange` determines that flush is needed, then `FlushOptions::wait` needs to be true, otherwise, it seems a violation of the contract of `CompactRange()`. It would be...

Jenkins run: https://jenkins.arangodb.biz/view/PR/job/arangodb-matrix-pr/23134/