Troubleshooting

Failures from the first build of this lab.

On this page

SSM RunShellScript is dash

AWS-RunShellScript is sh (dash on Ubuntu), not bash. set -o pipefail and bash arrays fail. Use aws ssm start-session and a bash shell, or wrap as echo BASE64 | base64 -d | bash.

gsm and apt prompt

gsm install without -y waits for confirmation. seiscomp install-deps calls apt install without -y. Use gsm install -y and wrap apt as in the walkthrough.

No outbound internet

The instance must have a public IPv4 (subnet map-public-ip plus --associate-public-ip-address, or an Elastic IP). There is no NAT. Without it, apt, gsm, SSM extras, and GEOFON all fail.

SSM not online

The instance profile must exist before launch. Wait about a minute after boot for PingStatus=Online.

seiscomp as root

seiscomp refuses root unless you pass --asroot. Run as sysop.

update-config deletes station keys

Inventory must exist before station_GE_* files. Missing stations (for example GE.BFO not in the FDSN response) are dropped.

xrdp drops after password

Globals port=3389 is the listen port. [Xorg] must be port=-1. A blanket sed on port= sets Xorg to 3389; xrdp then connects to itself.

Also check sesman.ini ListenAddress=0.0.0.0 and Xwrapper allowed_users=anybody.

scheli exits immediately

ERROR: no streams given — set heli.streams in ~/seiscomp/etc/scheli.cfg.

notify-send does nothing

Use /home/sysop/bin/sc-toast-event (gdbus against xfce4-notifyd). notify-send often hits xdg-portal and never shows.

RDP from Windows fails

Confirm your current public IP is the /32 on the security group. Choose session Xorg. Do not RDP to a Linux hop first.

PowerShell SSM port forwarding: use portNumber=3389,localPortNumber=13389, not JSON.

Elastic IP after stop

An EIP associated with a stopped instance still bills. Release it or keep the instance running.