Support Graviton instances and switch to EC2 fleets (4dn-dcic#375)

* Initial version * Minor changes * Some refactoring, add tests * Modifiy docs * Cleanup + version bump * Add AMIs * Address reviewer comments * Remove unnecessary permissions
drio18 · Dec 6, 2022 · e968c0c · e968c0c
1 parent b8e50bd
commit e968c0c
Show file tree

Hide file tree

Showing 30 changed files with 590 additions and 488 deletions.
diff --git a/CHANGELOG.rst b/CHANGELOG.rst
@@ -3,6 +3,15 @@
 Change Log
 ==========
 
+3.0.0
+=====
+
+* Added support for Graviton instances. 
+* Removed ``other_instance_types`` as option for ``behavior_on_capacity_limit``. It will fall back to ``wait_and_retry``.
+* Multiple instance types can be specified in the configuration. If ``spot_instance`` is enabled, Tibanna will run the workflow on the instance with the highest available capacity. If ``spot_instance`` is disabled, it will run the workflow on the cheapest instance in the list.
+* Instead of using the ``run_instance`` command we switch to EC2 fleets (in instant mode) to start up instances. 
+
+
 2.2.6
 =====
 

diff --git a/awsf3-docker/Dockerfile b/awsf3-docker/Dockerfile
@@ -35,25 +35,31 @@ RUN apt update -y && apt upgrade -y &&  apt install -y \
     libseccomp-dev \
     pkg-config \
     openjdk-8-jre-headless \
-    nodejs
+    nodejs \
+    gnupg \
+    lsb-release
 
 RUN ln -s /usr/bin/python3.8 /usr/bin/python
 #RUN ln -s /usr/bin/pip3 /usr/bin/pip
 
 WORKDIR /usr/local/bin
 
-# docker inside docker
-RUN curl -fsSL https://download.docker.com/linux/ubuntu/gpg | apt-key add - \
-    && apt-key fingerprint 0EBFCD88 \
-    && add-apt-repository "deb [arch=amd64] https://download.docker.com/linux/ubuntu $(lsb_release -cs) stable"
-RUN apt-get update -y \
-    && apt-cache policy docker-ce \
-    && apt-get install -y docker-ce
+# install docker inside docker
+RUN mkdir -p /etc/apt/keyrings
+RUN curl -fsSL https://download.docker.com/linux/ubuntu/gpg | gpg --dearmor -o /etc/apt/keyrings/docker.gpg
+
+RUN echo \
+  "deb [arch=$(dpkg --print-architecture) signed-by=/etc/apt/keyrings/docker.gpg] https://download.docker.com/linux/ubuntu \
+  $(lsb_release -cs) stable" | tee /etc/apt/sources.list.d/docker.list > /dev/null
+
+RUN apt-get update
+RUN apt-get --assume-yes install docker-ce
 
 # singularity
-RUN wget https://golang.org/dl/go1.16.6.linux-amd64.tar.gz && \
-    tar -xzf go1.16.6.linux-amd64.tar.gz && \
-    rm go1.16.6.linux-amd64.tar.gz
+RUN ARCH="$(dpkg --print-architecture)" && \
+    wget "https://golang.org/dl/go1.16.6.linux-${ARCH}.tar.gz" && \
+    tar -xzf "go1.16.6.linux-${ARCH}.tar.gz" && \
+    rm "go1.16.6.linux-${ARCH}.tar.gz"
 RUN export SINGULARITY_VERSION=3.8.1 && \
     export PATH=/usr/local/bin/go/bin/:$PATH && \
     wget https://github.com/sylabs/singularity/releases/download/v${SINGULARITY_VERSION}/singularity-ce-${SINGULARITY_VERSION}.tar.gz && \

diff --git a/awsf3-docker/run.sh b/awsf3-docker/run.sh
@@ -48,6 +48,7 @@ export TOPLATESTFILE=$LOCAL_OUTDIR/$JOBID.top_latest  # this one includes only t
 export INSTANCE_ID=$(ec2metadata --instance-id|cut -d' ' -f2)
 export INSTANCE_REGION=$(ec2metadata --availability-zone | sed 's/[a-z]$//')
 export INSTANCE_AVAILABILITY_ZONE=$(ec2metadata --availability-zone)
+export INSTANCE_TYPE=$(ec2metadata --instance-type)
 export AWS_ACCOUNT_ID=$(aws sts get-caller-identity| grep Account | sed 's/[^0-9]//g')
 export AWS_REGION=$INSTANCE_REGION  # this is for importing awsf3 package which imports tibanna package
 export LOCAL_OUTDIR_CWL=$MOUNT_DIR_PREFIX$LOCAL_OUTDIR

diff --git a/awsf3/aws_run_workflow_generic.sh b/awsf3/aws_run_workflow_generic.sh
@@ -182,8 +182,11 @@ exl echo "## Installing and activating Cloudwatch agent to collect metrics"
 cwd0=$(pwd)
 cd ~
 
+ARCHITECTURE="$(dpkg --print-architecture)"
+CW_AGENT_LINK="https://s3.amazonaws.com/amazoncloudwatch-agent/ubuntu/${ARCHITECTURE}/latest/amazon-cloudwatch-agent.deb"
 apt install -y wget
-wget https://s3.amazonaws.com/amazoncloudwatch-agent/ubuntu/amd64/latest/amazon-cloudwatch-agent.deb
+exl echo "Loading Cloudwatch Agent from ${CW_AGENT_LINK}"
+wget "${CW_AGENT_LINK}"
 sudo dpkg -i -E ./amazon-cloudwatch-agent.deb
 # If we want to collect new metrics, the following file has to be modified
 exl echo "## Using CW Agent config: https://raw.githubusercontent.com/4dn-dcic/tibanna/master/awsf3/cloudwatch_agent_config.json"

diff --git a/awsf3/cloudwatch_agent_config.json b/awsf3/cloudwatch_agent_config.json
@@ -42,7 +42,8 @@
 				"measurement": [
 					"io_time",
                     "read_bytes",
-                    "iops_in_progress"
+                    "iops_in_progress",
+					"read_time"
 				],
 				"metrics_collection_interval": 60,
 				"resources": [

diff --git a/awsf3/utils.py b/awsf3/utils.py
@@ -348,6 +348,7 @@ def update_postrun_json_init(json_old, json_new):
     prj.Job.instance_id = os.getenv('INSTANCE_ID')
     prj.Job.filesystem = os.getenv('EBS_DEVICE')
     prj.Job.instance_availablity_zone = os.getenv('INSTANCE_AVAILABILITY_ZONE')
+    prj.Job.instance_type = os.getenv('INSTANCE_TYPE')
 
     # write to new json file
     write_postrun_json(json_new, prj)

diff --git a/docs/ami.rst b/docs/ami.rst
@@ -2,7 +2,7 @@
 Amazon Machine Image
 ====================
 
-Tibanna now uses a single Amazon Machine Image (AMI) ``ami-06e2266f85063aabc``, which is made public for ``us-east-1``. One can find them among Community AMIs. (Tibanna automatically finds and uses them, so no need to worry about it.)
+Tibanna now uses the Amazon Machine Images (AMI) ``ami-06e2266f85063aabc`` (``x86``) and ``ami-0f3e90ad8e76c7a32`` (``Arm``), which are made public for ``us-east-1``. One can find them among Community AMIs. (Tibanna automatically finds and uses them, so no need to worry about it.)
 
-For regions that are not ``us-east-1``, a copy of the same AMI is publicly available (different AMI ID) and is auto-detected by Tibanna.
+For regions that are not ``us-east-1``, copies of these AMIs are publicly available (different AMI IDs) and are auto-detected by Tibanna.
 
diff --git a/docs/execution_json.rst b/docs/execution_json.rst
@@ -469,7 +469,12 @@ The ``config`` field describes execution configuration.
 :instance_type:
     - <instance_type>
     - This or ``mem`` and ``cpu`` are required if Benchmark is not available for a given workflow.
-    - If both ``instance_type`` and ``mem`` & ``cpu`` are specified, then ``instance_type`` is the first choice.
+    - ``instance_type`` can be a string (e.g.,  ``t3.micro``) or a list (e.g., ``[t3.micro, t3.small]``). If ``spot_instance``
+      is enabled, Tibanna will run the workflow on the instance with the highest available capacity. If ``spot_instance``
+      is disabled, it will run the workflow on the cheapest instance in the list.
+    - If both ``instance_type`` and ``mem`` & ``cpu`` are specified, Tibanna internally creates a list of instances that
+      are directly specified in ``instance_type`` and instances that satisfy the ``mem`` & ``cpu`` requirement. One instance is chosen
+      according to the rules above to run the workflow.
 
 :mem:
     - <memory_in_gb>
@@ -588,9 +593,7 @@ The ``config`` field describes execution configuration.
     - available options :
 
       - ``fail`` (default)
-      - ``wait_and_retry`` (wait and retry with the same instance type again),
-      - ``other_instance_types`` top 10 cost-effective instance types will be tried in the order
-                                 (``mem`` and ``cpu`` must be set in order for this to work),
+      - ``wait_and_retry`` (wait and retry with the same instance type again.),
       - ``retry_without_spot`` (try with the same instance type but not a spot instance) : this option is applicable only when
         ``spot_instance`` is set to ```True``
 

diff --git a/docs/news.rst b/docs/news.rst
@@ -19,6 +19,12 @@ Version updates
 
 .. _releases: https://github.com/4dn-dcic/tibanna/releases
 
+  **Nov 18, 2022** The latest version is now 3.0.0_.
+    - Tibanna now supports AWS Graviton-based instances.
+    - The instance type configuration now allows single instances (e.g., ``t3.micro``) and lists (e.g., ``[t3.micro, t3.small]``). If ``spot_instance`` is enabled, Tibanna will run the workflow on the instance with the highest available capacity. If ``spot_instance`` is disabled, it will run the workflow on the cheapest instance in the list.
+    - The option ``other_instance_types`` for ``behavior_on_capacity_limit`` has been removed. It will fall back to ``wait_and_retry``.
+
+
   **Mar 10, 2022** The latest version is now 2.0.0_.
     - The default Python version for Tibanna is now 3.8 (or 3.7). Python 3.6 is no longer supported.
 

diff --git a/poetry.lock b/poetry.lock
diff --git a/pyproject.toml b/pyproject.toml
@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "tibanna"
-version = "2.2.6"
+version = "3.0.0"
 description = "Tibanna runs portable pipelines (in CWL/WDL) on the AWS Cloud."
 authors = ["4DN-DCIC Team <[email protected]>"]
 license = "MIT"
@@ -43,7 +43,6 @@ flake8 = "^3.9.0"
 pytest = "^6.0"
 pytest-cov = "^3.0.0"
 pytest-parallel = "^0.1.1"
-mock = "4.0"
 pytest-mock = "3.7"
 
 coverage = {extras = ["toml"], version = "^6.3.2"}

diff --git a/scripts/publish-docker b/scripts/publish-docker
@@ -4,5 +4,6 @@
 export BUILD_LOG=/tmp/build-log
 export VERSION=$(python -c 'from tibanna._version import __version__; print(__version__)')
 export AWSF_IMAGE=$(python -c 'from tibanna.vars import DEFAULT_AWSF_IMAGE; print(DEFAULT_AWSF_IMAGE)')
-docker build -t $AWSF_IMAGE --build-arg version=$VERSION awsf3-docker/ > $BUILD_LOG
-docker push $AWSF_IMAGE
+# Your local docker driver needs to support the multiple platforms feature
+docker buildx build --push --platform linux/amd64,linux/arm64 -t $AWSF_IMAGE --build-arg version=$VERSION awsf3-docker/ > $BUILD_LOG
+
diff --git a/test_json/unicorn/medium_nonspot.postrun.json b/test_json/unicorn/medium_nonspot.postrun.json
@@ -53,6 +53,7 @@
         "status": "0",
         "filesystem": "/dev/nvme1n1",
         "instance_id": "i-01769a822e5dbb407",
+        "instance_type": "t3.medium",
         "instance_availablity_zone": "us-east-1b",
         "total_input_size": "12K",
         "total_output_size": "36K",

diff --git a/test_json/unicorn/small_spot.postrun.json b/test_json/unicorn/small_spot.postrun.json
@@ -53,6 +53,7 @@
         "status": "0",
         "filesystem": "/dev/nvme1n1",
         "instance_id": "i-01769a822e5dbb407",
+        "instance_type": "t3.small",
         "instance_availablity_zone": "us-east-1b",
         "total_input_size": "12K",
         "total_output_size": "36K",

diff --git a/test_json/unicorn/small_spot_gp3_iops.postrun.json b/test_json/unicorn/small_spot_gp3_iops.postrun.json
@@ -53,6 +53,7 @@
         "status": "0",
         "filesystem": "/dev/nvme1n1",
         "instance_id": "i-01769a822e5dbb407",
+        "instance_type": "t3.small",
         "instance_availablity_zone": "us-east-1b",
         "total_input_size": "12K",
         "total_output_size": "36K",

diff --git a/test_json/unicorn/small_spot_gp3_iops_throughput.postrun.json b/test_json/unicorn/small_spot_gp3_iops_throughput.postrun.json
@@ -53,6 +53,7 @@
         "status": "0",
         "filesystem": "/dev/nvme1n1",
         "instance_id": "i-01769a822e5dbb407",
+        "instance_type": "t3.small",
         "instance_availablity_zone": "us-east-1b",
         "total_input_size": "12K",
         "total_output_size": "36K",

diff --git a/test_json/unicorn/small_spot_io2_iops.postrun.json b/test_json/unicorn/small_spot_io2_iops.postrun.json
@@ -53,6 +53,7 @@
         "status": "0",
         "filesystem": "/dev/nvme1n1",
         "instance_id": "i-01769a822e5dbb407",
+        "instance_type": "t3.small",
         "instance_availablity_zone": "us-east-1f",
         "total_input_size": "12K",
         "total_output_size": "36K",

diff --git a/tests/awsf3/postrunjson/GBPtlqb2rFGH.postrun.json b/tests/awsf3/postrunjson/GBPtlqb2rFGH.postrun.json
@@ -112,6 +112,7 @@
         "filesystem": "",
         "instance_availablity_zone": "",
         "instance_id": "",
+        "instance_type": "",
         "start_time": "20210312-14:04:23-UTC"
     },
     "config": {