Project

General

Profile

Actions

Feature #63665

open

Feature #63663: mds,client: add crash-consistent snapshot support

mds: QuiesceDb to manage subvolume quiesce state

Added by Patrick Donnelly 6 months ago. Updated 2 months ago.

Status:
Pending Backport
Priority:
High
Assignee:
Category:
-
Target version:
% Done:

100%

Source:
Development
Tags:
backport_processed
Backport:
squid
Reviewed:
Affected Versions:
Component(FS):
MDS
Labels (FS):
Pull request ID:

Description

QuiesceDb is an ephemeral versioned replicated database holding a collection of QuiesceSet s. Each quiesce set carries a QuiesceState which can be observed by clients. QuiesceSet manages a collection of members - QuiesceRoot s and their corresponding QuiesceState s. When all members of a set reach quiesced state then the set is considered quiesced.

QuiesceDb is replicated across ranks of a file system. A single rank is selected to be the leader. The leader is authorized to maintain the database by processing external requests, peer acks, and timeout events. Updates to the database are sent out by the leader to all registered peers.

QuiesceDbManager is the class that maintains the quiesce db. Managers should be instantiated by the ranks and configured to form a cluster. The QuiesceDbManager only cares about the consistency of the quiesce db, and relies on the QuiesceAgent to perform the quiescing operations.


Related issues 1 (0 open1 closed)

Copied to CephFS - Backport #64896: squid: mds: QuiesceDb to manage subvolume quiesce stateResolvedLeonid UsovActions
Actions

Also available in: Atom PDF