CPU usage spike during long copy

[edgecode]

While copying a folder with over 130,000 ~50 MiB files from NAS to a client the CPU usage started to spike to 100% every few seconds on two cores (out of twenty). The usage was evenly distributed in single digits across all cores for the first several hours. The copy speed falls sharply during the spike.

TrueNAS-13.0-U5.3
Supermicro X10SRL-F
Xeon E5-2630 v4
64 GiB RAM
Single 10 GbE link (can't see the NIC model)

AFP share

 

[anodos]

While copying a folder with over 130,000 ~50 MiB files from NAS to a client the CPU usage started to spike to 100% every few seconds on two cores (out of twenty). The usage was evenly distributed in single digits across all cores for the first several hours. The copy speed falls sharply during the spike.

TrueNAS-13.0-U5.3
Supermicro X10SRL-F
Xeon E5-2630 v4
64 GiB RAM
Single 10 GbE link (can't see the NIC model)

AFP share

If all files are in a single directory, then this may indicate that client is periodically re-scanning the directory. IIRC netatalk is single-threaded and so this will correspond with CPU spike (because dir listing is often CPU bound) and then things will go back to normal after dir listing complete.

 

[edgecode]

So it is likely client side then? I see.
Thanks.

 

[anodos]

So it is likely client side then? I see.
Thanks.

Server response to client behavior is the way these things typically roll.