Scylla University Live
FREE Virtual Training Event | April 22, 2021

Register Now >

See all blog posts

Scylla release: version 1.6 RC1

The Scylla team is pleased to announce the release of Scylla 1.6 RC1, the first Release Candidate toward Scylla 1.6. From now on only critical bugs, if any, will be fixed in branch-1.6. We will continue to fix bugs and add features on the master branch.

Scylla is an open source, Apache-Cassandra-compatible NoSQL database, with superior performance and consistently low latency.Find Scylla 1.6 repository for your Linux distro here. Use with caution; RC1 is not officially production-ready yet. Help us stabilize 1.6 by reporting bugs here

Please note only the last two minor releases of the Scylla open source project are supported. Once 1.6 is out, 1.4 and earlier versions will not be supported, and there will be no maintenance release for them.

Scylla 1.6 focused on stability and bug fixes.

Noteworthy features

  • Debian Support (Experimental): starting from Scylla 1.6 we will provide Debian 8 packages. Debian repo will be shared soon.
  • Upgradable AMI. You can now use standard procedure to upgrade Scylla AMI version between Scylla maintenance, minor and major releases #1872, #1788.

Noteworthy new and updated metrics in Scylla 1.6

  • Starting in 1.6, metrics format is changed from seastar_(group){measurement="measurement", type="type"} to scylla_(group)_(measurement)_(type)
    If you are using scylla-grafana-monitoring, make sure to pull the latest release, and use 1.6 dashboards #1964
  • storage_proxy metrics names changed as follows:
    • Spaces in the metrics names have been replaced with underscores (“_”).
    • Per-DC/local Node split metrics names have changed:
      <name> (local Node) -> <name>_local_node
      <name> (external Nodes in DC: <DC name>) -> <name>_<DC name>
    • All split metrics have changed their type from total_operations to derive
      e.g. collectd metrics “total_operations-total write attempts (local Node)” became “derive-total_write_attempts_local_node”
    • Instead of a single storage_proxy group, there are two groups:
      1. storage_proxy_coordinator: counters updated by a Coordinator Node, e.g. read_timeouts. These include all but the storage_proxy_replica counters.
      2. storage_proxy_replica: counters updated by a Replica Node, e.g. received_mutations. These includes: received_mutationsforwarded_mutations, forwarding_errors

Noteworthy bug fixes

  • Improv data streaming used in repairs and out-scale #883, #1917, #1826, #1815 and others.
  • TitanDB does not work with Scylladb #693
  • Reduce bloom filter overhead in compaction #1322
  • scylla_setup: select disks to setup XFS showing only unmount disks #1841
  • scylla_setup: Scylla setup does not list local SSD disks on Google Cloud VM, or any other server with a device different from `/dev/sd*` `/dev/vd*` or `/dev/xvd*` #1963
  • Scylla can abort when running out of disk space #1764
  • Error in Scylla install of scylla user already exist #1958
  • Tracing: Starting a Node while other Nodes send RPC messages requesting tracing causes an assert #1939
  • Commit log replay, after restart, may drop Cell mutations of non-current (older) schema #1924
  • Writes are blocked for many seconds when Scylla is not able to keep up with flushing memtables #1879
  • Coredump may cause “no space left on device” on Ubuntu/Debian #1916
  • Scylla falls silently loaded on Ubuntu 16.04 #1199
  • Writes are blocked for many seconds when Scylla is not able to keep up with flushing memtables #1879
  • Scylla becomes very slow and eventually OOMs during large batch writes #1756
  • Thrift: Cell limit being applied across rows #1882
  • When Scylla restarts with large amount of data on disk, it can take a very long time until Scylla starts to accept requests #1856 (also part of 1.5)
  • When the ‘nodetool compact’ command is issued, in some circumstances, sstable file descriptors may stay open after deletion (by compaction), causing the disk space to run out #1840 (also part of 1.5)
  • When a scylla cluster of 2 or more is created, simultaneous creation of system_traces keyspace may trigger a family UUID disagreement between nodes #1976
Scylla Team

About Scylla Team

Scylla is the world’s fastest wide-column store database: the functionality of Apache Cassandra with the speed of a light key/value store.