-
ghuter authored
- add prefixing to tests outputs (test filenames) - use printf instead of echo in more places, for readability - redirect tests stderr to stdout - update readme to reflect changes - add an "examples" section to the readme
ghuter authored- add prefixing to tests outputs (test filenames) - use printf instead of echo in more places, for readability - redirect tests stderr to stdout - update readme to reflect changes - add an "examples" section to the readme
README 2.61 KiB
This directory is for tests against the g5k (grid5000) infrastructure. It should contain the "run" script, "ssh.pre.config", test executables, and a configuration file named "config". CONFIGURATION -------------- The configuration file contains records (lines) of colon-separated fields. These fields are, in order: a g5k site, a g5k cluster (or machine name) and a test executable filename. All blank lines, or lines beginning with a '#' shall be ignored. For instance, the following record specifies that the executable "test-load.sh" should be run on the "lille" g5k site and on the "chifflet" g5k cluster: lille:chifflet:test-load.sh As another example, the following example specifies that the executable "test-infiniband.sh" should be run on the "nancy" site and on the "grisou-49" machine (from the "grisou" cluster): nancy:grisou-49:test-infiniband.sh RUN --- "run" is an helper script to run specified tests against the g5k infrastructure. Usage: run -u <user> -i <identity_file> [<test_name_re> ...] where "user" is your g5k's login name, and "identity_file" would usually be something like "~/.ssh/grid5000.id_rsa" (in other words, this your g5k's private key). By default, without any other arguments, it will run all tests defined in the config file. If you specify any further arguments, it will interpret them as regular expressions and look for matches through the config file. Test results are sent back as their standard output and standard error output combined, each line prefixed with: "<test_filename>:". Therefore, one can have the output of, say, "test-load.sh", by piping `run` into `grep '^test-load.sh' | cut -d ':' -f 2-`. EXAMPLES -------- In the following examples, the given example config is supposed: ``` # site:cluster:test-filename lille:chifflet:test-nvidia.sh lille:chifflet:test-rapl.sh lille:chifflet:test-counters.sh lille:chifflet:test-temp.sh lille:chifflet:test-load.sh lille:chifflet:test-network.sh lille:chiclet:test-amd-rapl.sh lille:chiclet:test-infiniband.sh nancy:grisou-49:test-infiniband.sh ``` Run "test-load" test: ./run -u joe -i idfile test-load Run "test-counters" and "test-rapl" tests (notice that we write "test-rapl" in the command and not simply "rapl", because if we don't, both "test-rapl" and "test-amd-rapl" would be executed): ./run -u joe -i idfile counters test-rapl Run all lille tests: ./run -u joe -i idfile lille Run infiniband test on nancy site: ./run -u joe -i idfile 'nancy.*infiniband' Run infiniband test on all specified sites: ./run -u joe -i idfile infiniband Run all tests which are defined to use the chiclet cluster: ./run -u joe -i idfile chiclet