Parallel replay method for redo logs of file system

Abstract

The invention discloses a parallel replay method for redo logs of a file system. The parallel replay method comprises the following steps: step 1, a log reading thread is used for reading redo log entries in sequence from the redo logs of the file system and adding the redo log entries to an update action linked list; step 2, the linked list is scanned by multiple replay worker threads according to the sequence, if unprocessed redo log entries exist, each replay worker thread takes one redo log entry according to the sequence, and whether the redo log entry can be subjected to replay execution is judged on the basis of the principle for guaranteeing correct logical relationship between update actions; if the redo log entries can be subjected to replay execution, the replay worker threads replay the redo log entries, otherwise, the redo log entries are suspended; step 3, the replay worker threads finishing replay judge whether update actions waiting for finishing of metadata update actions exist, if yes, the replay worker threads wake up corresponding replay worker threads to scan the linked list again according to the sequence; otherwise, the linked list is scanned directly. The efficiency for replaying the redo logs of the file system is increased effectively.

Description

technical field [0001] The present invention relates to the field of computer technology. More specifically, it relates to a method for replaying file system redo logs in parallel. Background technique [0002] Computer file systems widely use redo log (redo log) technology in different fields to achieve various purposes such as file system data recovery, file system remote mirroring, and file system metadata replication. File system write log (WAL) / attempt log (Intent Long) is a form of redo log. In order to ensure the consistency of file system data and metadata, the file system will update the file system before the update operation is actually applied to the file system. All file system modifications involved in this update are persistently appended to the file system in the form of log entries. When these updates are actually applied to the file system, the relevant log entries can be recycled by the system. If the file system goes offline abnormally due to various re...

AI Technical Summary

Problems solved by technology

[0005] When the file system changes frequently, and the redo log client replays redo log entries faster than the log append speed (such as replaying the log to a remote mirror site), the redo log will become longer and longer, occupying More and more storage space

And slow replay of redo logs can affect the performance of file system redo log client applications

[0006] If the file system has a large number of small files and intensive metadata update operations (creation, deletion, modification operations), then for the redo log system, single-threaded sequential replay of redo log entries one by one cannot make full use of the computer Parallel resources, so inefficient, increased the burden on the redo log system

Parallel replay of redo logs is not easy to achieve, because there are dependencies between multiple metadata update actions in the file system

Images