fbpx

Scylla Manager 2.6

The Scylla Manager team is pleased to announce the release of Scylla Manager 2.6, a production-ready version of Scylla Manager for Scylla Enterprise customers and Scylla Open Source users. Scylla Manager is a centralized cluster administration and recurrent tasks automation tool.

Scylla Manager 2.6 is focused on gradual UX and stability improvements.

Scylla Enterprise customers are encouraged to upgrade to Scylla Manager 2.6 in coordination with the Scylla support team.

The new release includes upgrades of both Scylla Manager Server and Agent.

Useful Links:

Health-check Improvements

Manager Health-check periodically checks each managed node CQL/Rest availability (for Cassandra Query Language or Alternator, respectively). It is helpful if the Scylla node “thinks” it is up, but it is not available to the outside world.

Improvements to health-check in this release are:

New timeout configuration

It proves to be hard to set the timeout for Health-check, especially for multi-region clusters.

Starting in this release, one can specify the relative_timeout duration over the median time of a successful ping in a DC.

# Health-check service configuration.
#healthcheck:
# relative_timeout specifies timeout over median ping duration in probes.
# The number of probes kept in memory is specified by the probes parameter.
# There are separate probes for different DCs and ping types
# (CQL, REST, Alternator).
#  relative_timeout: 50ms
# max_timeout specifies maximum ping timeout, zero means no limit.
#  max_timeout: 30s
# Probes specifies how many probes are kept in memory for calculation.
# For different ping types and datacenters there are different probe sets.
#  probes: 200

Health Check Performance improvement

CQL query ping (when you provide credentials) is more efficient. It opens just a single connection and avoids initializing the whole driver stack.

Repair

A new sctool repair flag, --ignore-down-hosts, skips nodes that are down at the beginning of repair,  making managed repair run faster, not wasting time connecting to failed nodes.

Example:

$ sctool repair --dry-run -c test --ignore-down-hosts --host 192.168.100.11
NOTICE: dry run mode, repair is not scheduled

Host: 192.168.100.11

Ignore Hosts:
- 192.168.100.12
- 192.168.100.22

Data Centers:
- dc1
- dc2

Keyspaces:
- system_auth (3 tables)
- system_distributed (3 tables)
- system_traces (5 tables)
- test_keyspace_dc1_rf2 (1 table)
- test_keyspace_dc1_rf3 (1 table)
- test_keyspace_dc2_rf2 (1 table)
- test_keyspace_dc2_rf3 (1 table)
- test_keyspace_rf2 (1 table)
- test_keyspace_rf3 (1 table)

Suspend and Resume

sctool suspend command stops execution of all tasks that are running on a cluster. The tasks can be resumed using the resume command. If you want to stop a specific task, use the task stop command.

These commands are useful to implement a “Maintenance Windows“ making sure no manager task is running.

In this release  sctools suspend / resume commands made idempotent, calling suspend on an already suspended cluster or resuming twice would not produce errors.

sctools suspend / resume commands are idempotent, calling suspend on an already suspended cluster or resuming twice would not produce errors.

Monitoring

A new Scylla Manager agent metrics report on rclone errors.

04 Jan 2022