Farming cluster

[nazar-pc]

New build is cooking: Snapshot build · subspace/subspace@25ab72c · GitHub

It should improve situation with NATS-related messaging resulting in acknowledgement/ping errors, this is achieved by limiting piece retrieval concurrency to 100 pieces at any time and is also correlated to size of the connections pool to NATS (reduced from 16 in previous build to 4).

It is also possible to run multiple components in the same process if desired by separating them with -- (all will share the same pool of connections to NATS as well). Example of running cache and plotter together:

subspace-farmer cluster --nats-server nats://IP:4222 \
    cache \
        path=/path/to/cache,size=SIZE
    --
    plotter

---

@vexr

Plotter

Plotter warnings are te same as before and should be resolved in new build.

@vexr

Controller

This is not cluster-specific and rarely happens, nothing to worry about.

@vexr

16 * INFO async_nats: event: connected displayed at the start of each. If threads, perhaps combine them, or number them to show why the same message is being repeated.

There is a pool of connections to NATS now and it is configurable with corresponding CLI option, though I do not expect that you’ll need to change it.

@vexr

If not implemented, bumping replotting thread pools size on Plotter to match plotting thread pools size as default would make sense if the idea is to better consolidate resources with clustering.

There are no replotting CLI options in farming cluster plotter because plotter doesn’t know or care if request is for initial plotting or replotting. By default plotter will use all CPU cores fully all the time.

Let me know if you have a strong reason to not like this behavior, but expectation is that you will have some dedicated resources for plotting purposes (either separate machine or dedicated set of CPU cores for this purpose on machine used for other purposes too).