in the local network, latency is low
Your shares are always slow. Periodic plots enumeration is not the cause of this delay. Check your nfs setup. There is a bottleneck there. Huge network latency or busy hard drives. You may run fio benchmark tool to test it. Proof read requires dozens of sequential read operations. A 100ms latency can ruin the miner performance.
not only latency, IO and bandwidth have a great impactćminute scan all the plots Take up all my bandwidth, proof scan only scan some plots file
minute scan will cause faster damage to the hard disk also.
when a plot have many qualities you can read parallel not sequential.
Local network bandwidth should be more than enough for plot enumeration, which is done in parallel to proof read operations anyway and does not slow them down.
Everything is cached, even hard drive itself has an internal cache, it does not perform actual reading and does not receive any damage.
To remove stale shares you should decrease your nfs latency. It looks like you have 100-200ms latency now for each read operation. Its too much. If you are connecting to a distant storage with great latency you may consider running client remotely.
I have two nfs shares connected to one mining pc by 1gb network - no troubles. If I have stale, it happens then I uploads new plot to this nfs shares.
how many plot do you have? i have 8000 plot one machine.
1 Like
2500 plot per PC now
Some test stat - cache size and write to tmp file:
4gb ram = 908Gb write to tmp file
48gb ram = 396Gb write
56gb ram = 322Gb write
96gb ram = 50Gb write
106gb ram = 19Gb write
111gb ram = 9Gb write
113gb ram = 5Gb write
115gb ram = 2Gb write
116gb ram = 0,2Gb write
117gb ram = 0,09Gb write
1 Like
Ok. Iām just running the client with finalization and no plotting. Still crashing. Can someone point me in a direction to auto restart the client if it is not currently running? Iāve seen a million different ways of doing it but none of them are clicking with me. My additional ~30% gain will mean nothing if Iām down 33% of the time.
You can run it in docker and use its auto restart feature.
There must be a reason for crashes and it is probably not related to the client. Plotter is using lots of ram and may cause OOM, but not miner.
I have 175GiB allocated and finalizing/mining only uses a like 4. I have plotting tuned off.
Is this mean, more RAM and less HDD ??
Yes - more ram, less write to hdd\ssd, more speed.
2 Likes
I can help u get started in proxmox if u likeā¦
You solo farm.
Would u be interested in trying something with your solo plots?
No plots will be harmed I assure you.
I installed Proxmox on baremetal to test its just like VMwareās ESXI
Plot idā¦
Plus, youd only get 1 reward per block you " won" , youd get that same 1 reward without all that unnecessary work, because the plot will have the same proofs within it no matter how many times you link to it.
So why botherā¦
Why keep beating this broken drum.
2 Likes
@jonesjr I am flagging your post as inappropriate, as you already have a thread created by you where all that nonsense was discussed. So, if you have any new info that was requested you provide (e.g., you test this scenario and provide stats that support all those claims), you can update that thread, instead of trolling all other threads.
4 Likes