Refactor yacat example

[kmazurek]

Resolves #381

Key changes:

1. A single Golem instance is now used for executing the two phases of computation
2. Updated the CLI arguments for yacat

• added a default log location to examples utils so that the path doesn't need to be specified explicitly
• changed hash and mask to be named, required arguments (instead of positional ones)
• replaced --number-of-providers with --chunk-size
• added optional --max-workers argument to allow for manual override of the default dynamic value
• added optional --hash-type argument

3. Slimmed down the Dockerfile used for building the VM image by using a pre-built hashcat image

• updated the volume paths that are used in this image

4. Reduced the number of files generated on the requestor to temporary files used for storing the results from providers
5. Refactored both worker functions

[shadeofblue]

maybe DICTIONARY_ATTACK_TIMEOUT ? ... just "attack" seems more serious than it is to a casual (non-security-terminology-aware) user ...

[kmazurek]

Makes sense, though using the word dictionary in this case may be confusing (i.e. in hashcat dictionary and mask are two different types of attacks). How about MASK_ATTACK_TIMEOUT and renaming the worker funcion to perform_mask_attack?

[kmazurek]

Updated in daad9bc

[shadeofblue]

lgtm

just one minor remark

[azawlocki]

I'd provide an example value for --mask somewhere, preferably in help strings in the parser, to make this example runnable more easily without referring to external docs (apart from general golem/yapapi docs)

[azawlocki]

Or maybe provide an example invocation, with --mask and --hash, in the string passed to build_parser()?

[kmazurek]

Good one! Actually I might just do both. 🤷

[kmazurek]

Added in a0b544c

[azawlocki]

No need to check if cmd_result.success == True here, if it weren't then the line future_result = yield context.commit(timeout=KEYSPACE_TIMEOUT) would raise yapapi.executor.CommandExecutionError.

[azawlocki]

I'd rewrite the body of compute_keyspace() as follows:

    async for task in tasks:
        cmd = f"xhashcat --keyspace " f"-a {HASHCAT_ATTACK_MODE} -m {args.hash_type} {args.mask}"
        context.run("/bin/bash", "-c", cmd)

        try:
            future_result = yield context.commit(timeout=KEYSPACE_TIMEOUT)
            # each item is the result of a single command on the provider (including setup commands)
            result: List[CommandExecuted] = await future_result
            # we take the last item since it's the last command that was executed on the provider
            cmd_result: CommandExecuted = result[-1]
            keyspace = int(cmd_result.stdout)
            task.accept_result(result=keyspace)
        except CommandExecutionError as e:
            raise RuntimeError(f"Failed to compute attack keyspace: {e}")

[azawlocki]

That would print for example:

[2021-05-27 13:07:04,963 WARNING yapapi.summary] Worker for provider 'provider-2' failed; reason: Failed to compute attack keyspace: Command '{'run': {'entry_point': '/bin/bash', 'args': ('-c', 'xhashcat --keyspace -a 3 -m 400 ?a?a?a'), 'capture': {'stdout': {'stream': {}}, 'stderr': {'stream': {}}}}}' failed on provider; message: 'ExeScript command exited with code 127'; stderr: 'bash: xhashcat: command not found

[kmazurek]

Good point, I'll switch to a try-except in compute_keyspace.

[kmazurek]

Changed in 2ab9620

[azawlocki]

It's a much nicer read than the original!

I'd consider changing the body of compute_keyspace() to handle CommandExecutionError instead of checking the status of the returned CommandExecuted event. Or we can just let the exception bubble up without wrapping it in a RuntimeError.

[azawlocki]

LGTM!