ZFS on the server and the desktop: N+1 ways to better store your data

Presented on 21 November 2013 at the NLUUG fall conference in Bunnik, The Netherlands

Transcript

1. ZFS on the server and the desktop N+1 ways to better store your data Presented at NLUUG 21-11-2013
2. About Matthias van der Heide Storage Network Engineer Inprove IT BV an EraStor member @matthiasvdh on Twitter
3. Contents • ZFS Intro • Features • N+1 ways • Why use it in DevOps • Where to get it
4. What is ZFS? • “Zettabyte File System” • Sun, now Oracle • Thank Sun for OpenSolaris and releasing ZFS
5. ZFS Features • Yes, it is a file system • It is also volume manager • Virtually unlimited capacity • Transactional • Multiple types of redundancy • Virtually unlimited snapshots, clones • Smart caching • Versioned on-disk format
6. Copy-on-write
7. Checksums NL 08 IN GB 00 00 00 05 55
8. ARC • Adaptive Replacement Cache • Combines MRU and MFU • Uses ghost lists • Consumes (a lot of) memory, extend to SSD
9. ZIL • ZFS Intent Log • In memory • On the pool • Optionally on a seperate log device (SSD)
10. Transaction Groups (TXG) Open • Accept new writes Quiescing • Wait for I/O to complete Syncing • Write the data to disk
11. N+1 ways • Designing your pool • Testing your pool • Tuning the filesystem • Maintenance
12. N=1: Designing your pool • Your performance and redundancy needs determine your pool design • You can change it, only slightly, afterwards • Consider a system with 4 HDDs • Now what do I do?
13. Redundancy types • A vdev is the building block of the zpool • Everything is striped over vdevs • Options: – Leaf (1) – Mirror (2..n) – RAID-Z1 (n+1) – RAID-Z2 (n+2)
14. Options • Stripe over all drives • Stripe over mirrors • RAID-Z1 • RAID-Z2
15. Fun options • Dual RAID-Z1 • 4-way mirror • HDD as slog device
16. N=2: Testing your pool • Simple: time dd • Iometer • Vdbench • FIO
17. N=3: Tuning your filesystem • Block size • Log bias • Access time • Sync on/off • Caching on/off
18. N=4: Maintenance • ZFS will tell you when stuff is broken • Check data on all reads • Regular scrubs
19. Scrubs • Scans the pool for data integrity • Checks checksums • Restores data if needed • Can take a while on large pools
20. Scrub
21. Scrub time 0,00 10,00 20,00 30,00 40,00 50,00 60,00 70,00 80,00 90,00 100,00 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 Estimate accuracy Estimate accuracy
22. Resilvers • When replacing a device • When adding a device back after removal
23. Simple detach/attach
24. After being offline
25. New device
26. Future work on ZFS • Feature flags enables development of features seperately • Persistent L2ARC • Larger block support • Dedupe tables on SSD
27. Why use it in DevOps • ZFS performs • ZFS scales – From your laptop to your workstation to your cloud • ZFS replicates – Using zfs send/receive you can deploy copies of your VMs to other hosts – Keep simple backups of your repos and machines • Use snapshots to boot into new versions of your operating system
28. Where to get it? • FreeBSD • Mac • ZFS on Linux • Illumos (EraStor OS) • NexentaStor (based on OpenSolaris/Illumos) • OpenZFS - http://open-zfs.org
29. Links • Open ZFS Project http://open-zfs.org • ARC Paper http://citeseer.ist.psu.edu/viewdoc/summary?doi=10.1.1.1 3.5210 • NOSIG http://www.nosig.nl/ • Erastor http://www.erastor.eu