Hello all, I started to talk about this issue in another topic, but I think it needs a space on its own.

my system is an Hp server xeon 2630v3 32gig ddr4 11 disks with dual parity and ssd cache pool with two sata ssd in raid 1, dirty ram cache ratio 20,60.

I just configured a 10gbe network with some issues, lastly writing to Ram cache was slower as expected. Thanks to the forum and some digging I found out having this SHFS issue, which is bypassed by writing to cache disk directly.

During this long troubleshooting I found out that while writing a 10gb file to the server cache enabled share, it first went to the ram taking 10 seconds (circa 1GB/s) and than flushed to the ssd disk in 20 seconds (circa 500MB/s), for a total 30seconds. The speed are normal for the hardware, but to my understanding it should fill up the ram, ending the file transfer, and than flushing to cache in background, thus allowing for other tasks or transfer to be done. What I observe is that the file transfer window stays stucked at the end until the whole file is flushed to the cache pool, thus making the file transfer globally Slower.

is this an Osx (Mojave 10.14) related issue or should I tune something to “disengage” the file transfer?

It kills the whole purpose of ram caching.

 

Edited May 21, 20215 yr by mo679

Not a mac user but IIRC there was an issue with osx that treated SMB writes as sync writes, google "osx smb strict sync".

Thanks for pointing me in the right direction, I changed settings as showed in the guide and other posts here in the forum, now it works without hanging.

next..Ram upgrade

Hallo again, I changed computer and moved to a M1 macbook, using Big Sur. and the hanging of the files has started again, i changed the config on the server side before in smb confing, and that was fine, 

should I change also something on the Mac, I tried to change the strict sync parameters but it does not change anything in Big Sur