ScyllaDB Documentation Logo Documentation
  • Server
    • ScyllaDB Open Source
    • ScyllaDB Enterprise
    • ScyllaDB Alternator
  • Cloud
  • Tools
    • ScyllaDB Manager
    • ScyllaDB Monitoring Stack
    • ScyllaDB Operator
  • Drivers
    • CQL Drivers
    • DynamoDB Drivers
Download
Menu
Scylla Manager Repair Repair slower

Repair slower¶

You can make repair run slower by changing the level of parallelism or intensity. By default Scylla Manager runs repairs with full parallelism and intensity one.

Reduce amount of work performed by repair master node. Try setting --intensity 0.5, that reduces the number of token ranges repaired by a repair master node in a single chunk of work. You can further decrease it to almost 0 forcing Scylla to repair a single token range at a time. Beware that 0 is a special value, it means max intensity.

Try setting --parallel 1, that would cap the number of Scylla repair jobs in the cluster to one, frees nodes that are not being repaired. This would have the same result as running Scylla Manager 2.1 or earlier in terms of parallelism.

Slowdown a running repair¶

If a repair is running on a cluster, you can change intensity and parallelism levels while it is running. Run the following command to repair less token ranges on nodes at the same time. In the following example the number of token ranges equals half of the number of shards of a node.

sctool repair control -c prod-cluster --intensity 0.5

Run the following command to limit the number of parallel Scylla repair jobs in the cluster to one.

sctool repair control -c prod-cluster --parallel 1

You can view your current values for intensity and parallel with the sctool progress command. Stopping the task and running again, would reset the values.

Slowdown the future runs of a repair¶

If you wish to change intensity and parallelism level of a repair task use sctool repair update command. Run the following command to repair less token ranges on nodes at the same time. In the following example the number of token ranges equals half of the number of shards of a node.

sctool repair update -c prod-cluster repair/143d160f-e53c-4890-a9e7-149561376cfd --intensity 0.5

Run the following command to limit the number of parallel Scylla repair jobs in the cluster to one.

sctool repair update -c prod-cluster repair/143d160f-e53c-4890-a9e7-149561376cfd --parallel 1
PREVIOUS
Repair faster
NEXT
Examples
  • 3.0
    • master
    • 3.0
    • 2.6
    • 2.5
    • 2.4
    • 2.3
    • 2.2
  • Docker
  • Install Scylla Manager
  • Install Scylla Manager Agent
  • Scylla Manager Upgrade
  • Add a Cluster
  • Backup
    • Setup AWS S3
    • Setup S3 compatible storage
    • Setup Google Cloud Storage
    • Setup Azure Blob Storage
    • Examples
    • Specification
  • Restore
    • Download files command
  • Repair
    • Repair faster
    • Repair slower
    • Examples
  • Health Check
  • CLI sctool
    • Environment variables
    • Completion
    • Backup
    • Cluster
    • Info
    • Progress
    • Repair
    • Start
    • Status
    • Stop
    • Suspend & Resume
    • Tasks
    • Version
  • Configuration Files
    • Scylla Manager Config
    • Scylla Manager Agent Config
  • Swagger File
  • Scylla Monitoring
  • Troubleshooting
  • Older Versions
  • Slack
  • Create an issue
  • Edit this page

On this page

  • Repair slower
    • Slowdown a running repair
    • Slowdown the future runs of a repair
Logo
Docs Contact Us About Us
Mail List Icon Slack Icon Forum Icon
© 2023, ScyllaDB. All rights reserved.
Last updated on 15 February 2023.
Powered by Sphinx 4.3.2 & ScyllaDB Theme 1.3.4