atlasdb
atlasdb copied to clipboard
palantir
Transactional Distributed Database Layer
## Before this PR AtlasDB uses linked lists in a couple places. I noticed this while analyzing a heap dump from an OOM of one of our highest volume services...
## General **Before this PR**: `ImmutableCassandraServer` instances are created and used as Set elements & Map keys heavily, and would benefit from canonicalizing as there should be a relatively small,...
## General **Before this PR**: AtlasDB CLI could fail if KVS async initialization is enabled, _and_ the KVS is used before it is initialized. This is because use throws a...
**Goals (and why)**: Teams using AtlasDB can diagnose service-level statistics, but trace-level usage information needs to be reconstructed indirectly. This PR is the first step to create easy-to-digest statistics to...
## General **Before this PR**: Runtime config was not supported by the migration command in the AtlasDB CLI. As services migrate over to specifying servers and replicationFactor in the runtime...
## General **Before this PR**: It isn't actually feasible to wire up atlasdb proxy to autobatch its unlock requests **After this PR**: ==COMMIT_MSG== Unlock requests across multiple namespaces in atlasdb...
## General **Before this PR**: There was no way to mark a transaction in progress. **After this PR**: Introduces `markInProgress` in `TransactionService` for transactions4. Does not perform any wiring. The...
## General **Before this PR**: All of these exceptions were simply RTEs **After this PR**: We have a new exception that we can declare to slightly reduce the jank level...
## General **Before this PR**: There doesn't exist storage for the set *A*, which I call the futile timestamps store **After this PR**: ==COMMIT_MSG== There exists a store for *A*...