when running piece cache sync on cluster, after a while it pauses for indefinite time
Environment
amd 7950X3D
windows 11
subspace-farmer-windows-x86_64-skylake-direct-io-linux-macos-backport
all cluster components and node on one machine
the issue is reproducible on every attempt.
the link to the log files:
(I hope it was actually reproduced, since I couldn’t understand what’s going on with all that output. After a while I noticed that network usage has dropped significantly so I presumed the process has paused)
I think the problem here is that connection to cache dropped somehow (at least on logical level, like with timeouts during plotting). However, piece cache sync is unaware of this and continues the process facing a bunch of errors in the process, which in turn likely prevents cache from being re-discovered and reinitialized.
I would like to ask, when my cache component is planned with two spaces, do these two spaces contain the same cache data, or is the cache evenly distributed to the two spaces?
Each cache group contains its own distinct cache data, separate from the others. As long as each cache group is equal to or larger than the “Archived History Size” of the blockchain, you will have two complete caches based on your example once fully synced.
If you sync one cache group first, you can stop the cache component and copy the fully synced cache file to speed up the syncing of the second group, effectively reducing the total sync time by half as they contain the same data.