You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Situation: I have my storages mounted on my daemon (for the windowpost) through NFS. I don't want the finalized sealed sector to go from worker -> daemon -> storage machine, so I created a lotus worker on the storage machine with the storage attached.
Lotus walks through the entire sealing process, and ends up with s-1 doing a GET for the sealed sector.
Daemon thinks "oh its on my remote w-c2 machine" so S-1 fetches from http://daemon:2345/remote (stupid, because why not just fetch from http://w-c2:2345/remote directly?)
Storage has it fetched and stored on disk. Now tells the remote to delete the sealed sector (because they don't need it anymore).
Daemon thinks "oh yeah I have that one local, in the NFS mount, lol" and deletes it.
Endpoint: FinalizeSector is stuck and all sealed data (complete sector) is gone.
I know I can prevent this by removing the s-1 worker. But then the daemon would do all the fetching directly into its own mounted NFS drive, effectively making the data do a stupid loop from W-c2 -> daemon -> S-1.
Preventing either one would be best.
The text was updated successfully, but these errors were encountered:
Lotus component
Lotus Version
Describe the Bug
Situation: I have my storages mounted on my daemon (for the windowpost) through NFS. I don't want the finalized sealed sector to go from worker -> daemon -> storage machine, so I created a lotus worker on the storage machine with the storage attached.
So, my s-1 storage has 2 URL's:
Server: daemon (runs lotus daemon and miner)
Worker: w-pc1
Worker: w-pc2
Worker: w-c2
Storage: s-1
Lotus walks through the entire sealing process, and ends up with s-1 doing a GET for the sealed sector.
Daemon thinks "oh its on my remote w-c2 machine" so S-1 fetches from http://daemon:2345/remote (stupid, because why not just fetch from http://w-c2:2345/remote directly?)
Storage has it fetched and stored on disk. Now tells the remote to delete the sealed sector (because they don't need it anymore).
Daemon thinks "oh yeah I have that one local, in the NFS mount, lol" and deletes it.
Endpoint: FinalizeSector is stuck and all sealed data (complete sector) is gone.
I know I can prevent this by removing the s-1 worker. But then the daemon would do all the fetching directly into its own mounted NFS drive, effectively making the data do a stupid loop from W-c2 -> daemon -> S-1.
Preventing either one would be best.
The text was updated successfully, but these errors were encountered: