Feature #9255

[keep] drain mode for a keepstore service

Added by Nico C├ęsar over 5 years ago. Updated almost 2 years ago.

Assigned To:
Target version:
Start date:
Due date:
% Done:


Estimated time:
Story points:


I'd like to have a way to tell a keepstore service that it will enter in drain mode. This will cause to switch itself to read-only mode and then check for under replicated blocks that it owns and send copies of it to the other servers

features to have:
  • a way to tell the progress of a draining operation and if it has finished or not.
  • if a VM reboot happens of the keepstore service, the operation should not start from scratch
  • we known for a fact that multi-threaded copies work very good, but could also kill the machine or I/O operattions: this should be an option on how many processes will be spawning for the migration

things to take in account:

  • if 7 keepstore services are in drain state, replication should be taking in account for the NON-draining ones, this will be common operation for migrations between clouds.
  • clients should be aware of the draining state and no take in account for replication calculations.

Related issues

Related to Arvados - Feature #14714: [keep] keep-balance uses cluster config fileResolved09/26/2019


#1 Updated by Tom Clegg over 2 years ago

  • Related to Feature #14714: [keep] keep-balance uses cluster config file added

Also available in: Atom PDF