Hello, Everyone.
On our site, a server with 16 hyper-threading cores managed 168 disks, each disk configured with one file system.
Currently, the scanrate and scan_ns_rate values for individual disks are set to 50 MB/s and 100 by default, which I believe is too high for a single server.
I was wondering if other sites or CERN use these numbers and if you have any suggestions.
I’d be grateful if you could share your experience.
Regards,
– Geonmo
Hi Geonmo,
In one of our production instances that also holds a sizable amount of RAIN files we use 10 MB/s as scanrate and the default 50 files/s for scan_ns_rate. The effect of lowering to these values is that the scanning will take longer, but you can easily monitor this in your system and check how long the ScanDir takes on average per disk by looking for messages like the one below, in the systemctl status output of the FST daemons:
Jul 30 16:28:51 st-048-a02076a6.cern.ch scandir[1767481]: [ScanDir] Directory: /data03 files=0 scanduration=0 [s] scansize=0 [Bytes] [ 0 MB ] scannedfiles=0 corruptedfiles=0 hwcorrupted=0 skippedfiles=0 disk_scan_interval_sec=14400
Cheers,
Elvin