vitrivr / cthulhu Goto Github PK

We have to be cautious ; )

For instance, a BashJob can only be executed in a unix worker. It would be useful to define requirements for a job to be able to run properly in a worker. (not all workers are created equal : )

The current implementation logic assumes that a job that is being updated, and had not been sent to a Worker is an inconsistency and an error. Is this reasonable? (Maybe yes)

Issue

As seen in #21 IPv6 addresses would break the Schedulers communication with the worker that gave the IP address. This has been circumvented by the worker not giving their IPv6 address.

Resolution

Add the ability to parse IPv6 addresses to the scheduler and allow the worker to use them.

Is it necessary for the default constructors of Job and BashJob to be public? Don't they just have to exist in order for the reflection to work?

The type of a job should probably be represented using an enum, since there will be a fixed set of existing types.

Package names should start with the 'org.vitrivr.cthulhu' prefix. Package names of other vitrivr projects will be adjusted accordingly once the repositories are moved into the vitrivr organization.

This routine needs to be strengthened:

• Job deletion must be conditional on state of the job (receive a boolean force parameter)
• Return value must be useful

What should a worker do when its coordinator becomes unreachable/stops responding?

It would be great if the static files for the UI could be bundled into the jar and be served from there.

What is the idea behind the static final ints JOB_FAILURE, JOB_INTERRUPTION, etc.? Is the plan to or them together as a set of status bits or are these independent states? In case they are independent, I would again go for an enum.

Maybe it should also consider other factors (job type? job creation time?)

See in index.html, and find the following lines:

<script src="https://ajax.googleapis.com/ajax/libs/jquery/2.2.4/jquery.min.js"></script>
<link rel="stylesheet" href="http://yui.yahooapis.com/pure/0.6.0/pure-min.css"></link>
<script src="http://warfares.github.io/pretty-json/pretty-json-min.js" type="text/javascript"></script>

Perhaps it should be able to deploy in machines without internet access.

This makes the worker stop fetching jobs from its queue. Currently this is not a problem because jobs don't keep several jobs in queue; but it could be.

It doesn't right now.

Gotta handle those errors

that's it

It'd be great : )