Hi,
I have tested the boundary-reference-architecture/README.md at main · hashicorp/boundary-reference-architecture · GitHub repository for AWS deployment.
After creating both controllers and workers ubuntu servers on my AWS account I have connected to one of the controllers to check if the install.sh was copied to the machine but when trying to run it (using the same command on terraform “sudo ~/./install.sh controller” and it write:
: not foundtall.sh: 4:
also from terraform itself we can see that remote-exec: not found pops up:
I have checked with simple script install2.sh “hello world” to check if there is some problem with the commands on thee server but it runs fine
what do I need to proceed with running the install.sh script as it should?
I just tried the reference architecture and it worked as it should, e.g. this happens on one of my controllers at the provisioning stage:
module.aws.aws_instance.controller[1] (remote-exec): Connected!
module.aws.aws_instance.controller[1] (remote-exec): Adding system user `boundary' (UID 113) ...
module.aws.aws_instance.controller[1] (remote-exec): Adding new group `boundary' (GID 119) ...
module.aws.aws_instance.controller[1] (remote-exec): Adding new user `boundary' (UID 113) with group `boundary' ...
module.aws.aws_instance.controller[1] (remote-exec): Creating home directory `/home/boundary' ...
module.aws.aws_instance.controller[1] (remote-exec): Unable to capture a lock on the database.
module.aws.aws_instance.controller[1] (remote-exec): Created symlink /etc/systemd/system/multi-user.target.wants/boundary-controller.service → /etc/systemd/system/boundary-controller.service.
module.aws.aws_instance.controller[1]: Creation complete after 3m18s [id=*redacted*]
…which is exactly what should happen.
Has the install.sh script maybe gotten changed? The sha256sum of the copy on my controller is:
$ sha256sum /home/ubuntu/install.sh
0b844f64a04dabcf313e23c4381d3396f326fa656362ce5b8e220ea3c9cf7c06 /home/ubuntu/install.sh
Yes It is, can you provide me the exact install.sh? I haven’t changed there nothing
Installs the boundary as a service for systemd on linux
TYPE=$1
NAME=boundary
sudo cat << EOF > /etc/systemd/system/{NAME}-{TYPE}.service
[Unit]
Description={NAME} {TYPE}
[Service]
ExecStart=/usr/local/bin/{NAME} server -config /etc/{NAME}-${TYPE}.hcl
User=boundary
Group=boundary
LimitMEMLOCK=infinity
Capabilities=CAP_IPC_LOCK+ep
CapabilityBoundingSet=CAP_SYSLOG CAP_IPC_LOCK
[Install]
WantedBy=multi-user.target
EOF
Add the boundary system user and group to ensure we have a no-login
user capable of owning and running Boundary
sudo adduser --system --group boundary || true
sudo chown boundary:boundary /etc/{NAME}-{TYPE}.hcl
sudo chown boundary:boundary /usr/local/bin/boundary
Make sure to initialize the DB before starting the service. This will result in
a database already initizalized warning if another controller or worker has done this
already, making it a lazy, best effort initialization
if [ “${TYPE}” = “controller” ]; then
sudo /usr/local/bin/boundary database init -skip-auth-method-creation -skip-host-resources-creation -skip-scopes-creation -skip-target-creation -config /etc/{NAME}-{TYPE}.hcl || true
fi
sudo chmod 664 /etc/systemd/system/{NAME}-{TYPE}.service
sudo systemctl daemon-reload
sudo systemctl enable {NAME}-{TYPE}
sudo systemctl start {NAME}-{TYPE}
####This is my install.sh
Also when trying to run manually commands from install.sh on DB init I get:“command not found”
Its from DB Init row in the script
Your script looks fine – I’m not sure why the sums are coming out differently for us. Might be as simple as a trailing newline or something. In any case the known-good version is here.
There are two things that might be going on here:
One is that the uploaded Boundary binary should be located at /usr/local/bin/boundary – there shouldn’t be a directory called that like there is in your screenshot, it should be the actual Boundary binary. Make sure your -var boundary_bin=[path to Boundary] points to the directory that contains the boundary binary within itself, not a higher-level directory.
The other is that if the Boundary binary you pointed at with boundary_bin is a Windows binary, it won’t work when uploaded to a Linux host. You’d need to download the Linux amd64 binary and point at that.
As a fellow Windows user, I ran into this same problem (actually both problems) so I thought it might be useful to document what happened for anyone else who stumbles across this.
When I cloned the boundary-reference-architecture repo, evidently something translated the files to windows format. All of the line endings show up as CRLF rather than the unix standard of LF. That doesn’t seem to be a problem for the *.tf files, but it does cause issues for the install.sh file.
Took me the better part of a day to figure that out…
