niksu / dosarray Goto Github PK

Specifically, automate this command: https://github.com/niksu/dosarray/blob/master/doc/SETUP.md#configuring-docker
Could do this by extending this script: https://github.com/niksu/dosarray/blob/master/src/dosarray_configure_networking.sh

Currently error messages are sometimes written to stdout (e.g., d87c55b#diff-4cf78e31b37364c390ea8c708e594d19R110) but this can make it more difficult to hunt down why the system failed.

Feature involves easier terminal access to all containers from the controller.
Currently it's possible to easily access the hosts from the controller, but sometimes it's useful to access the containers to run ping etc from them to run some manual tests, unlike the scripted tests in the experiments.

Currently load graphing involves several manual steps (described here) but this could easily be automated so that the data files for load measurements (for CPU load, memory and network) are all produced, and subsequently the graphs (for all kinds of load) are generated.

e.g., if instances have been created/started, and if an experiment might currently be running.
For the instances, it would report how many (out of the total number expected) are currently in the created/running state (in case something broke during an earlier run, and the system is only partly running.)

for each host in DOSARRAY_PHYSICAL_HOSTS_PUB check to see that the host responds, it's running docker, it has the image available, has networking configured, and that its network settings match what src/dosarray_configure_networking.sh would have configured.

This could be done by following this example: fcb21ec
Having such logs could help ease debugging.

Sometimes bits of experiments don't work correctly -- e.g., load testing of CPU or network.
The idea is to have a script that runs an post-experiment analysis to find out if anything looks wrong.
This is similar to this #25 -- except that it's run after an experiment, not before, and lints or looks for bad signs in that experiment's output.

add option in the experiment configuration to make it easier to generate multiple kinds of graph (from the same data, from that experiment); currently this is done manually.

Info on motivation+usage of https://github.com/niksu/dosarray/blob/bb4f0a980829a1c360563ca6e3990bcc66b596c5/src/dosarray_synch_time.sh could be added to https://github.com/niksu/dosarray/blob/bb4f0a980829a1c360563ca6e3990bcc66b596c5/doc/USAGE.md

This script would reverse all the configuration made by DoSarray to a cluster, returning it to a previous state. It would complement the scripts to stop and delete containers.
This script would be useful to return the cluster into a state that others can use without any risk of interference from DoSarray's configuration (even if DoSarray isn't currently running)