I was trying to figure out why SMBD was chugging and maxing out a CPU core. It was still maxing my local LAN (1Gbps) for 1-2 transfers, but if multiple clients hit the server I'd see performance drop and SMBD was obviously the bottleneck, not the disk array.
The problem turned out to be so simple that I never noticed it. Turns out I'd somehow/somewhen (I have no recollection of doing this) set the log level for the SMBD service to "debug" instead of "normal". After I changed that from "debug" to "normal" my CPU usage fell from ~90% to 20%.
Note that this is on a Xeon CPU - an older one, but still not an Atom or something basic. I was shocked at how much of an impact that setting change made.
Just something to check for if you see SMBD bottlenecking.
The problem turned out to be so simple that I never noticed it. Turns out I'd somehow/somewhen (I have no recollection of doing this) set the log level for the SMBD service to "debug" instead of "normal". After I changed that from "debug" to "normal" my CPU usage fell from ~90% to 20%.
Note that this is on a Xeon CPU - an older one, but still not an Atom or something basic. I was shocked at how much of an impact that setting change made.
Just something to check for if you see SMBD bottlenecking.
