xrootd / xrootd-k8s-operator Goto Github PK

Idea 1

• Create a top-level client-facing proxy for xrootd protocol communication. It would find and help the client connect to the correct worker using public IP DNS record.

Idea 2

• Create Custom K8S endpoints to map worker hostnames to public IPs

Idea 3 (IDEAL)

• Expose workers using NodePort headless service and hope the client gets the correct worker IP.

Common options to configure in xrootd configuration, xrootd.cf

This will allow other operators, like https://github.com/lsst/qserv-operator, to benefit from reusing the xrootd cluster in their qserv cluster, since most of the xrootd fundamentals are duplicated in both the operators.

Shivansh, it seems your integration test scripts continue even if it met an error.
CI script must not continue on error, so having a 0 return code at the end of CI mean it is ok. Having a non-zero code mean it has not passed. Currently some commands crash during the CI, but the CI script returns 0, this must not occur.

I recommend you use
"set -e" so that all you shell scripts crashes on error, this is much safer.

I would also recommend to use
"set -euxo pipefail" with bash, as recommended by Google good practice, and try to stay the more possible POSIX compliant.

Here:

Could you also use a simple shell variable instead of 'set'?
TEST_SCRIPTS=$(ls $DIR/../tests/e2e/test-*.sh)
I think this look much simpler.

An finally, thanks for all your work on the xrootd-operator, this is great!