When using NATS Server+Controller+Farmer+Plot method. If the controller cannot synchronize the cache 100%. So starting plotter is invalid. Must wait for 100% completion of cache. Start the plotter separately again. in other words. Starting the plot first cannot automatically wait for the cache to complete. Start the plot directly. It needs to be manually started
I want to clarify the behaviour you’re seeing.
When you start controller, cache, farmer, plotter all at the same time we would expect the cache to initialise and potentially sync so that it picks up any new pieces of history.
Are you saying that the plotter waits for the cache synchronisation before starting plotting and replotting? But that restarting the plotter will have it start re/plotting even while the cache is syncing?
It is impossible to start them simultaneously. I started in sequence: node ->nats server ->cluster ->controller ->cache ->farmer ->plotter. However, if the cache is not 100% complete. So starting plotter won’t work. We need to wait for cahe to complete 100%. Start plotter separately again. Otherwise, even if cahe synchronizes at any time. Ploters are not working.
This sounds like you want the controller to notify the farmer when the cache sync is complete so it can ask the plotter to start?
If the cache is not synced you could see plotting going very slow as it has to reach out to the DSN for the pieces not in the cache. We just discussed a slow piece retrieval issue on Taurus on the Dev Sync. Can you provide logs so we can see exactly what is happening please? Probably the controller, farmer and plotter would cover it.
You still haven’t fully understood. If plotter is started before cache 100% synchronization. It is a useless startup (it will remain stuck there, doing nothing). Only after 100% cache synchronization. Restarting plotter is the effective start. It just works.
./subspace-farmer cluster --nats-server nats://127.0.0.1:4222 plotter \
2025-06-02T18:45:52.105768Z INFO async_nats: event: connected
2025-06-02T18:45:53.339070Z INFO subspace_farmer::commands::cluster::plotter: Using CUDA GPUs used_cuda_devices=[0]
2025-06-02T18:45:53.354927Z INFO subspace_farmer::commands::cluster::plotter: Multiple L3 cache groups detected l3_cache_groups=2
2025-06-02T18:45:53.354960Z INFO subspace_farmer::commands::cluster::plotter: Preparing plotting thread pools plotting_thread_pool_core_indices=[CpuCoreSet { cores: CpuSet(0-13,28-41), .. }, CpuCoreSet { cores: CpuSet(14-27,42-55), .. }]
2025-06-02T18:45:53.360338Z INFO subspace_farmer::commands::cluster::plotter: CPU plotting was disabled due to detected faster plotting with GPU
Just like a log. There is no longer any reaction. Even if there are new cache downloads. That’s also the case. in other words. In a NATS cluster. If the cache is not synchronized. Or restart the cache server (especially restart the cache). All plotters need to be restarted once. Otherwise, it would just get stuck there like this