| layout | title | permalink | redirect_from | ||
|---|---|---|---|---|---|
post |
GETTING STARTED |
/docs/getting-started |
|
AIStore runs on a single Linux or Mac machine. Bare-metal Kubernetes and GCP/GKE Cloud-based deployments are also supported. There are numerous other options.
Generally, when deciding how to deploy a system like AIS with so many possibilities to choose from, a good place to start would be answering the following two fundamental questions:
- what's the dataset size, or sizes?
- what hardware will I use?
For datasets, say, below 50TB a single host may suffice and should, therefore, be considered a viable option. On the other hand, the Cloud deployment option may sound attractive for its ubiquitous convenience and for not thinking about the hardware and the sizes - at least, not right away.
Note as well that you can always start small: a single-host deployment, a 3-node cluster in the Cloud or on-premises, etc. AIStore supports many options to inter-connect existing clusters - the capability called unified global namespace - or migrate existing datasets (on-demand or via supported storage services). For introductions and further pointers, please refer to the AIStore Overview.
AIStore runs on commodity Linux machines with no special requirements whatsoever. It is expected that within a given cluster, all AIS targets are identical, hardware-wise.
- Linux (with
GCC,sysstatandattrpackages, and kernel 4.15+) or macOS - Go 1.19 or later
- Extended attributes (
xattrs- see next section) - Optionally, Amazon (AWS), Google Cloud Platform (GCP), and/or Azure Cloud Storage accounts.
See also:
CROSS_COMPILEcomment below.
Depending on your Linux distribution, you may or may not have GCC, sysstat, and/or attr packages. These packages must be installed.
Speaking of distributions, our current default recommendation is Ubuntu Server 20.04 LTS. But Ubuntu 18.04 and CentOS 8.x (or later) will also work. As well as numerous others.
For the local filesystem, we currently recommend xfs. But again, this (default) recommendation shall not be interpreted as a limitation of any kind: other fine choices include zfs, ext4, f2fs, and more.
Since AIS itself provides n-way mirroring and erasure coding, hardware RAID would not be recommended. But can be used, and will work.
The capability called extended attributes, or xattrs, is a long-time POSIX legacy supported by all mainstream filesystems with no exceptions. Unfortunately, xattrs may not always be enabled in the Linux kernel configurations - the fact that can be easily found out by running the setfattr command.
If disabled, please make sure to enable xattrs in your Linux kernel configuration. To quickly check:
$ touch foo
$ setfattr -n user.bar -v ttt foo
$ getfattr -n user.bar foomacOS/Darwin is also supported, albeit for development only. Certain capabilities related to querying the state and status of local hardware resources (memory, CPU, disks) may be missing, which is why we strongly recommend Linux for production deployments.
The rest of this document is structured as follows:
- Local Playground
- Make
- Multiple deployment options
- Kubernetes Playground
- HTTPS
- Build, Make and Development Tools
- Containerized Deployments: Host Resource Sharing
For a quick evaluation, experimenting with features, first-time usage, and (of course) development - for any and all of the above running AIS from its GitHub source is maybe the first option to consider.
Hence, Local Playground - one of the several supported deployment options.
Local Playground is not intended for production and is not meant to provide optimal performance.
To run AIStore from source, you'd typically need Go: compiler, linker, tools, and required packages. However:
CROSS_COMPILEoption (see below) can be used to build AIStore without having (to install) Go and its toolchain.
To install Go(lang) on Linux:
- download the latest
go1.19.<x>.linux-amd64.tar.gzfrom Go downloads - follow installation instructions
- or simply run:
tar -C /usr/local -xzf go1.19.<x>.linux-amd64.tar.gz
Next, if not done yet, export the GOPATH environment variable.
Here's an additional 5-minute introduction that talks about setting up the Go environment and also includes:
- provisioning data drives for AIS deployment, and
- running a single-node AIS cluster locally.
Once done, run AIS as follows:
The steps:
$ cd $GOPATH/src/github.com/NVIDIA
$ git clone https://github.com/NVIDIA/aistore.git
$ cd aistore
# optionally, run `make mod-tidy` to preload dependencies
$ ./deploy/scripts/clean_deploy.sh
$ ais show clusterwhere:
clean_deploy.shwith no arguments builds AIStore binaries (such asaisnodeandaisCLI) and then deploys a local cluster with 5 proxies and 5 targets. Examples:
# Deploy 7 targets and 1 proxy:
$ clean_deploy.sh --proxy-cnt 1 --target-cnt 7
# Same as above, plus built-in support for GCP (cloud storage):
$ clean_deploy.sh --proxy-cnt 1 --target-cnt 7 --gcpFor more options and detailed descriptions, run make help and see: clean_deploy.sh.
Here's a quick, albeit somewhat outdated, YouTube introduction and demo.
AIStore (product and solution) is fully based on HTTP(S) utilizing the protocol both externally (to support both frontend interfaces and communications with remote backends) and internally, for intra-cluster streaming.
Connectivity-wise, what that means is that your local deployment at localhost:8080 can as easily run at any arbitrary HTTP(S) address.
Here're the quick change you make to deploy Local Playground at (e.g.) 10.0.0.207, whereby the main gateway's listening port would still remain 8080 default:
diff --git a/deploy/dev/local/aisnode_config.sh b/deploy/dev/local/aisnode_config.sh |
index 9198c0de4..be63f50d0 100755 |
--- a/deploy/dev/local/aisnode_config.sh |
+++ b/deploy/dev/local/aisnode_config.sh |
@@ -181,7 +181,7 @@ cat > $AIS_LOCAL_CONF_FILE <<EOL |
"confdir": "${AIS_CONF_DIR:-/etc/ais/}", |
"log_dir": "${AIS_LOG_DIR:-/tmp/ais$NEXT_TIER/log}", |
"host_net": { |
- "hostname": "${HOSTNAME_LIST}", |
+ "hostname": "10.0.0.207", |
"hostname_intra_control": "${HOSTNAME_LIST_INTRA_CONTROL}", |
"hostname_intra_data": "${HOSTNAME_LIST_INTRA_DATA}", |
"port": "${PORT:-8080}", |
diff --git a/deploy/dev/local/deploy.sh b/deploy/dev/local/deploy.sh |
index e0b467d82..b18361155 100755 |
--- a/deploy/dev/local/deploy.sh |
+++ b/deploy/dev/local/deploy.sh |
@@ -68,7 +68,7 @@ else |
PORT_INTRA_DATA=${PORT_INTRA_DATA:-13080} |
NEXT_TIER="_next" |
fi |
-AIS_PRIMARY_URL="http://localhost:$PORT" |
+AIS_PRIMARY_URL="http://10.0.0.207:$PORT" |
if $AIS_USE_HTTPS; then |
AIS_PRIMARY_URL="https://localhost:$PORT" |AIS comes with its own build system that we use to build both standalone binaries and container images for a variety of deployment options.
The very first make command you may want to execute could as well be:
$ make helpThis shows all subcommands, environment variables, and numerous usage examples, including:
Examples:
# Deploy cluster locally
$ make deploy
# Stop locally deployed cluster and cleanup all cluster-related data and bucket metadata (but not cluster map)
$ make kill clean
# Stop and then deploy (non-interactively) cluster consisting of 7 targets (4 mountpaths each) and 2 proxies; build `aisnode` executable with the support for GCP and AWS backends
$ make kill deploy <<< $'7\n2\n4\ny\ny\nn\nn\n0\n'All containerized deployments have their own separate Makefiles. With the exception of local playground, each specific build-able development (dev/) and production (prod/) option under the deploy folder has a pair: {Dockerfile, Makefile}.
This separation is typically small in size and easily readable and maintainable.
Also supported is the option not to have the required Go installed and configured. To still be able to build AIS binaries without Go on your machine, make sure that you have docker and simply uncomment CROSS_COMPILE line in the top Makefile.
AIStore deploys anywhere anytime supporting multiple deployment options summarized and further referenced here.
In particular:
For any Kubernetes deployments (including, of course, production deployments) please use a separate and dedicated AIS-K8s GitHub repository. The repo contains Helm Charts and detailed Playbooks that cover a variety of use cases and configurations.
In particular, AIS-K8s GitHub repository provides a single-line command to deploy Kubernetes cluster and the underlying infrastructure with the AIStore cluster running inside (see below). The only requirement is having a few dependencies preinstalled (in particular, helm) and a Cloud account.
The following GIF illustrates steps to deploy AIS on the Google Cloud Platform (GCP):
Finally, the repository hosts the Kubernetes Operator project that will eventually replace Helm charts and will become the main deployment, lifecycle, and operation management "vehicle" for AIStore.
This option has the unmatched convenience of requiring an absolute minimum time and resources - please see this README for details.
You can also run make deploy in the root directory of the repository to deploy a cluster:
$ make deploy
Enter number of storage targets:
10
Enter number of proxies (gateways):
3
Number of local cache directories (enter 0 to use preconfigured filesystems):
2
Select backend providers:
Amazon S3: (y/n) ?
n
Google Cloud Storage: (y/n) ?
n
Azure: (y/n) ?
n
HDFS: (y/n) ?
n
Create loopback devices (note that it may take some time): (y/n) ?
n
Building aisnode: version=df24df77 providers=Notice the "Cloud" prompt above and the fact that access to 3rd party Cloud storage is a deployment-time option.
Run make help for supported (make) options and usage examples, including:
# Restart a cluster of 7 targets (4 mountpaths each) and 2 proxies; utilize previously generated (pre-shutdown) local configurations
$ make restart <<< $'7\n2\n4\ny\ny\nn\nn\n0\n'
# Redeploy the cluster (4 targets, 1 proxyi, 4 mountoaths); build `aisnode` executable for debug without any backend-supporting libraries; use RUN_ARGS to pass an additional command-line option ('-override_backends=true') to each running node
$ RUN_ARGS=-override_backends MODE=debug make kill deploy <<< $'4\n1\n4\nn\nn\nn\nn\n0\n'
# Same as above, but additionally run all 4 targets in a standby mode
$ RUN_ARGS='-override_backends -standby' MODE=debug make kill deploy <<< $'4\n1\n4\nn\nn\nn\nn\n0\n'
...
...Further:
make kill- terminate local AIStore.make restart- shut it down and immediately restart using the existing configuration.make help- show make options and usage examples.
For even more development options and tools, please refer to:
For development, health-checking a new deployment, or for any other (functional and performance testing) related reason you can run any/all of the included tests.
For example:
$ go test ./ais/tests -v -run=MirrorThe go test above will create an AIS bucket, configure it as a two-way mirror, generate thousands of random objects, read them all several times, and then destroy the replicas and eventually the bucket as well.
Alternatively, if you happen to have Amazon and/or Google Cloud account, make sure to specify the corresponding (S3 or GCS) bucket name when running go test commands.
For example, the following will download objects from your (presumably) S3 bucket and distribute them across AIStore:
$ BUCKET=aws://myS3bucket go test ./ais/tests -v -run=downloadTo run all tests in the category short tests:
# using randomly named ais://nnn bucket (that will be created on the fly and destroyed in the end):
$ BUCKET=ais://nnn make test-short
# with existing Google Cloud bucket gs://myGCPbucket
$ BUCKET=gs://myGCPbucket make test-shortThe command randomly shuffles existing short tests and then, depending on your platform, usually takes anywhere between 15 and 30 minutes. To terminate, press Ctrl-C at any time.
Ctrl-C or any other (kind of) abnormal termination of a running test may have a side effect of leaving some test data in the test bucket.
In our development and testing, we make use of Minikube and the capability, further documented here, to run the Kubernetes cluster on a single development machine. There's a distinct advantage that AIStore extensions that require Kubernetes - such as Extract-Transform-Load, for example - can be developed rather efficiently.
In the end, all examples above run a bunch of local web servers that listen for plain HTTP requests. Following are quick steps for developers to engage HTTPS:
- Generate X.509 certificate:
$ openssl req -x509 -newkey rsa:4096 -keyout server.key -out server.crt -days 1080 -nodes -subj '/CN=localhost'- Deploy cluster (4 targets, 1 gateway, 6 mountpaths, Google Cloud):
$ AIS_USE_HTTPS=true AIS_SKIP_VERIFY_CRT=true make kill deploy <<< $'4\n1\n6\nn\ny\nn\nn\n0\n'- Run tests (both examples below list the names of buckets accessible for you in Google Cloud):
$ AIS_ENDPOINT=https://localhost:8080 AIS_SKIP_VERIFY_CRT=true BUCKET=gs://myGCPbucket go test -v -p 1 -count 1 ./ais/tests -run=ListBuckets
$ AIS_ENDPOINT=https://localhost:8080 AIS_SKIP_VERIFY_CRT=true BUCKET=tmp go test -v -p 1 -count 1 ./ais/tests -run=ListBucketsNotice environment variables above: AIS_USE_HTTPS, AIS_ENDPOINT, and AIS_SKIP_VERIFY_CRT.
As noted, the project utilizes GNU make to build and run things both locally and remotely (e.g., when deploying AIStore via Kubernetes. As the very first step, run make help for help on:
- building AIS binary (called
aisnode) deployable as both a storage target or a proxy/gateway; - building CLI, aisfs, and benchmark binaries;
In particular, the make provides a growing number of developer-friendly commands to:
- deploy the AIS cluster on your local development machine;
- run all or selected tests;
- instrument AIS binary with race detection, CPU and/or memory profiling, and more.
The following applies to all containerized deployments:
- AIS nodes always automatically detect containerization.
- If deployed as a container, each AIS node independently discovers whether its own container's memory and/or CPU resources are restricted.
- Finally, the node then abides by those restrictions.
To that end, each AIS node at startup loads and parses cgroup settings for the container and, if the number of CPUs is restricted, adjusts the number of allocated system threads for its goroutines.
This adjustment is accomplished via the Go runtime GOMAXPROCS variable. For in-depth information on CPU bandwidth control and scheduling in a multi-container environment, please refer to the CFS Bandwidth Control document.
Further, given the container's cgroup/memory limitation, each AIS node adjusts the amount of memory available for itself.
Memory limits may affect dSort performance forcing it to "spill" the content associated with in-progress resharding into local drives. The same is true for erasure-coding which also requires memory to rebuild objects from slices, etc.
For technical details on AIS memory management, please see this readme.