ZFS Internals July 2014
Wednesday, July 9, 2014 at 9:00 AM - Friday, July 11, 2014 at 5:00 PM (PDT)
San Francisco, CA
Duration: 3 Days
Trainer: Max Bruning
This course covers the internal implementation of the ZFS file system. It uses a combination of source code and various tools available on illumos to examine the topics that are covered. Tools used include: zdb, DTrace, and mdb. The course starts with a description of the ZFS on-disk format. It then moves to a description of how ZFS worksin the kernel. This will include examination of mounted ZFS file systems, how reads and writes work, and how storage is managed while it is being used. The course concludes with a brief look at some problems that may arise when using ZFS, and possible ways of dealing with these problems.
On completion of this course, students will be able to:
- Explain the layout of ZFS file systems on disk.
- Identify the metadata and data for a given file within a zpool.
- Describe the mechanics of mounting a ZFS dataset.
- Use zdb(1M) to examine the metadata for a ZFS dataset.
- Explain how ARC works.
- Explain the function of the DMU, transactions, and transaction groups.
- Describe ZFS transaction groups.
- Describe the ZIO pipeline
This course is designed for anyone who has a need to know how ZFS works. This could be a developer, system administrator, or support/maintenance personnel.
Good knowledge of illumos internals (could be Solaris internals), or the SmartOS Internals course.
Module 1: Introduction
- Overview of ZFS Components
Module 2: ZFS On-Disk Format
- Object Sets
- Space Maps
- Snapshots and Clones
- On disk layout
Module 3: ZFS in memory
- Mounted ZFS file systems
- ZFS Posix Layer
- ZFS Volumes
- Transaction Groups
- Reading file data
- Writing file data
- ZIO pipeline
Module 4: zdb
- Example uses of zdb to examine ZFS on disk format