Advanced

Changing the port used by gibbs

By default, gibbs uses the port 5019 for the sockets the communicate.

You can change this port by using the argument gibbs_port in the Worker constructor, and the argument port in the Hub constructor :

hub = Hub(port=6222)

w = Worker(MyModel, gibbs_port=6222)

Passing arguments to model's constructor

If your model requires arguments for the constructor (like in the example given in the Usage section) :

class MyAwesomeModel:
    def __init__(self, wait_time=0.25):
        super().__init__()
        self.w = wait_time

You can just pass the arguments to the Worker (positional arguments or keyword arguments, both work) :

# With positional argument
w1 = Worker(MyAwesomeModel, 0.3)

# With keyword argument
w2 = Worker(MyAwesomeModel, wait_time=0.3)

---

Only a small list of keywords arguments are reserved for gibbs. Here is the exhaustive list :

• gibbs_host
• gibbs_port
• gibbs_heartbeat_interval
• gibbs_reset_after_n_miss

Starting workers in another machine

By default, gibbs workers try to connect to a Hub on the local machine (localhost).

But you can change this behavior, to have workers running in another machine !

To do this, simply start your worker with the appropriate host using the gibbs_host argument :

w = Worker(MyModel, gibbs_host="192.178.0.3")

Retrials & timeouts

By default, the request() method of the Hub will indefinitely waits for a response from the worker.

---

If instead you want to have a timeout, you can specify the timeout (in seconds) with the gibbs_timeout argument :

h = Hub()
await h.request(x, gibbs_timeout=0.4)

If the Hub doesn't receive any response from the worker within the specified timeout, a asyncio.TimeoutError exception is raised.

---

You can also specify a number of retries with the argument gibbs_retries :

h = Hub()
await h.request(x, gibbs_timeout=2, gibbs_retries=3)

With this code, the Hub will try to send a request. If the Hub didn't receive an answer within 2 seconds, it will retry the request again, up to 3 times.

Note

You need to specify gibbs_timeout when using gibbs_retries, because there is no timeout by default !

---

You can also specify gibbs_retries=-1 for infinite retries !

Logging

Inside gibbs, the library loguru is used for logging.

By default, the logger is disabled (so logging functions become no-op). If you want to see what's going on inside gibbs (for contributing or debugging for example), you can activate it :

from loguru import logger

logger.enable("gibbs")
# From here, logs will be displayed

---

You can also change the level of logs you want, the format, etc...

from loguru import logger
import sys

logger.enable("gibbs")
logger.remove()
logger.add(sys.stderr, format="{time} {level} {message}", level="INFO")

Tip

For more details, check out the loguru library !

Changing the heartbeat interval

By default, the heartbeat interval is set to one second.

You can change this value by using the argument gibbs_heartbeat_interval in the Worker constructor, and the argument heartbeat_interval in the Hub constructor :

hub = Hub(heartbeat_interval=10)

w = Worker(MyModel, gibbs_heartbeat_interval=10)

Warning

Make sure that both the Hub and the workers have the same value for the heartbeat interval, otherwise they might have synchronisation issues !