ternfs-XTXMarkets/README.md

TernFS

A distributed file system. For a high-level description of TernFS, see its [blog post](TODO insert link). This document provides a more bare-bones overview and an introduction to the codebase.

Goals

The target use case for TernFS is the kind of machine learning we do at XTX: reading and writing large immutable files. With "immutable" we mean files that do not need modifying after they are first created. With "large" we mean that most of the storage space will be taken up by files bigger than a few MBs.

We don't expect new directories to be created often, and files (or directories) to be moved between directories often. In terms of numbers, we expect the upper bound for TernFS to roughly be the upper bound for the data we're planning for a single data center:

• 10EB of logical file storage (i.e. if you sum all file sizes = 10EB)
• 1 trillion files -- average ~10MB file size
• 100 billion directories -- average ~10 files per directory
• 1 million clients

We want to drive the filesystem with commodity hardware and Ethernet networking.

We want the system to be robust in various ways:

• Witnessing half-written files should be impossible -- a file is fully written by the writer or not readable by other clients
• Power loss or similar failure of storage or metadata nodes should not result in a corrupted filesystems (be it metadata or filesystem corruption)
• Corrupted reads due to hard drives bitrot should be exceedingly unlikely
• Data loss should be exceedingly unlikely, unless we suffer a datacenter-wide catastrophic event (fire, flooding, datacenter-wide vibration, etc)
• The filesystem should keep working through maintenance or failure of metadata or storage nodes

We also want to be able to restore deleted files or directories, using a configurable "permanent deletion" policy.

Finally, TernFS can be replicated to multiple data centres to make it resilient to data centre loss, with each data centre storing the whole TernFS data set.

Components

TODO decorate list below with links drilling down on specific concepts.

• servers
  • shuckle
    • 1 logical instance
    • ternshuckle, Go binary
    • state currently persisted through SQLite (1 physical instance), should move to a Galera cluster soon (see #41)
    • TCP -- both bincode and HTTP
    • stores metadata about a specific TernFS deployment
      • shard/cdc addresses
      • block services addressea and storage statistics
      • latency histograms
    • serves web UI
  • filesystem data
    • metadata
      • shard
        • 256 logical instances
        • ternshard, C++ binary
        • stores all metadata for the filesystem
          • file attributes (size, mtime, atime)
          • directory attributes (mtime)
          • directories listings (includes file/directory names)
          • file to blocks mapping
          • block service to file mapping
        • UDP bincode req/resp
        • state persisted through RocksDB, currently single physical instance per logical instance, will move to a Paxos (or similar) cluster per logical instance (see #56)
        • communicates with shuckle to fetch block services, register itself, insert statistics
    • CDC
      • 1 logical instance
      • terncdc, C++ binary
      • coordinates actions which span multiple directories
        • create directory
        • remove directory
        • move file or directory between from one directory to the other
        • "Cross Directory Coordinator"
      • UDP bincode req/resp
      • very little state required
        • information about which transactions are currently being executed and which are queued (currently transactions are executed serially)
        • directory -> parent directory mapping to perform "no loops" checks
      • state persisted through RocksDB, currently single physical instance, will move to Paxos or similar just like shards (see #56)
      • communicates with the shards to perform the cross-directory actions
      • communicates with shuckle to register itself, fetch shards, insert statistics
  • block service
    • up to 1 million logical instances
    • 1 logical instance = 1 disk
    • 1 physical instance handles ~100 logical instances (since there are ~100 disks per server)
    • ternblocks, Go binary (for now, will be rewritten in C++ eventually)
    • stores "blocks", which are blobs of data which contain file contents
    • each file is split in many blocks stored on many block services (so that if up to 4 block services fail we can always recover files)
    • single instances are not redundant, the redundancy is handled by spreading files over many instances so that we can recover their contents
    • TCP bincode req/resp
    • extremely dumb, the only state is the blobs themselves
    • its entire job is efficiently streaming blobs of data from disks into TCP connections
    • communicates with shuckle to register itself and to update information about free space, number of blocks, etc.
• clients, these all talk to all of the servers
  • cli
    • terncli, Go binary
    • toolkit to perform various tasks, most notably
      • migrating contents of dead block services (terncli migrate)
      • moving around blocks so that files are stored efficiently (terncli defrag, currently WIP, see #50)
  • kmod
    • ternfs.ko, C Linux kernel module
    • kernel module implementing mount -t eggsfs ...
    • the most fun and pleasant part of the codebase
  • FUSE
    • ternfuse, Go FUSE implementation of a TernFS client
    • much slower but should be almost fully functional (there are some limitations concerning when a file gets flushed)
• daemons, these also talk to all of the servers
  • GC
    • terngc, Go binary
    • permanently deletes expired snapshots (i.e. deleted but not yet purged data)
    • cleans up all blocks for permanently deleted files
  • scrubber
    • TODO see #32
    • goes around detecting and repairing bitrot
• additional tools
  • ternrun, a tool to quickly spin up a TernFS instance
  • terntests, runs some integration tests

Building

% ./build.sh alpine

Will build all the artifacts apart from the Kernel module. The output will be in build/alpine. Things will be built in an Alpine Linux container, so that everything will be fully statically linked.

There's also ./build.sh ubuntu which will do the same but in a Ubuntu container (we use this to build for production), and ./build.sh release which will build outside docker, which means that you'll have to install some dependencies in the host machine.

Testing

./ci.py --build --integration --short --docker

Will run the integration tests as CI would (inside a docker image). You can also run the tests outside docker by removing the --docker flag, but you might have to install some dependencies of the build process. These tests take a few minutes.

To work with the qemu kmod tests you'll first need to download the base Ubuntu image we use for testing:

% wget -P kmod 'https://cloud-images.ubuntu.com/focal/current/focal-server-cloudimg-amd64.img'

Then you can run the CI tests in kmod like so:

% ./ci.py --kmod --short --prepare-image=kmod/focal-server-cloudimg-amd64.img --leader-only

The tests redirect dmesg output to kmod/dmesg, event tracing output to kmod/trace, and the full test log to kmod/test-out.

You can also ssh into the qemu which is running the tests with

% ssh -p 2223 -i kmod/image-key fmazzol@localhost

Note that the kmod tests are very long (~1hr). Usually when developing the kernel module it's best to use ./kmod/restartsession.sh to be dropped into qemu, and then run specific tests using terntests.

However when merging code modifying TernFS internals it's very important for the kmod tests to pass as well as the normal integration tests. This is due to the fact that the qemu environment is generally very different from the non-qemu env, which means that sometimes it'll surface issues that the non-qemu environment won't.

Playing with a local TernFS instance

% cd go/ternrun
% go run . -data-dir <data-dir>

The above will run all the processes needed to run TernFS. This includes:

• 256 metadata shards;
• 1 cross directory coordinator (CDC)
• A bunch of block services (this is tunable with the -flash-block-services, -hdd-block-services, and -failure-domains flags)
• 1 shuckle instance

A multitude of directories to persist the whole thing will appear in <data-dir>. The filesystem will also be mounted using FUSE under <data-dir>/fuse/mnt.

TODO add instruction on how to run a multi-datacenter setup.

Building the Kernel module

% cd kmod
% ./fetchlinux.sh # fetch linux sources
% (cd linux && make oldconfig && make prepare && make -j) # build Linux
% make KDIR=linux -j kmod

VS Code

Most of the codebase is understandable by VS Code/LSP:

• Code in go/ just works out of the box with the Go extension. I (fmazzol) open a separate VS Code window which specifically has the ternfs/go directory open, since the Go extension doesn't seem to like working from a subdirectory.
• Code in cpp/:

  • Disable existing C++ integrations for VS Code (I don't remember which exact C++ extension caused me trouble -- something by Microsoft itself).

  • Install the clangd extension.

  • Generate compile_commands.json with

    % cd cpp
    % ./build.py debug
    % cp ./build/debug/compile_commands.json .
    

    If you change the build process (i.e. if you change CMake files) you'll need to generate and copy compile_commands.json again.

• Code in kmod/:
  • Build the module.
  • Generate compile_commands.json with ./kmod/gen_compile_commands.sh.
  • New files should work automatically, but if things stop working, just re-bulid and re-generate compile_commands.json.

A note on naming

TernFS was originally called EggsFS internally. This name quickly proved to be very poor due to the phonetic similarity to XFS, another notable filesystem. Therefore the filesystem was renamed to TernFS before open sourcing. However the old name lingers on in certain areas of the system that would have been hard to change, such as metric names.