On 2022-07-22 21:12, balin...@gmail.com wrote:
Hello Everyone,
I have 4 x300 (8 channels) recording 20MHz bandwidth continuously on
all 8 channels at the same time onto a RAID. That’s 645MB/s of data
rate recorded onto the RAID. Sometimes the recording goes on for 10
hours without any issues whatsoever…sometimes it hiccups every 4 hours
with /*“Overflow (Out of Sequence)*/ errors. Below is the system
specs. Continued after system specs…
—————————————————————————————-
*CPU:* AMD Ryzen 7 5700G 8: Cores 16 threads
*RAM*: 32GB RAM DD4
*Motherboard:* ASUS Pro WS x570-ACE with latest BIOS ver 4201
*Ethernet Card*: 4-Port Intel X710 10G Ethernet Card with i40e driver
as of March 2022 —> plugged into PCIe x8 and running at x8
*OS*: Linux Mint 20.3 Una 64 bit with Kernel 5.15.041 (Clean, fresh
install via reformat)
*OS Drive*: 1TB Samsung NVME EVO 970 Plus
*RAID:* Areca 1883IX-24: RAID 6 w/12x 16TB Seagate Exos (145TB ext4
partition) —> plugged into PCIe x8 and running at x8. Latest
Firmware/BIOS/Bootloader package dated June 2022.
——————————-
*UHD ver:* 3.15.0.0.rc3 (UHD_3.15.0.HEAD-0-g4e06022c)
4x x300 via 10G connection to the back of the Intel X710 4port 10G card.
*Ring buffers:* maxed out at RX 4096 and TX 4096 on the Intel X710 on
all 4 ports
*Memory buffers*: rmem and wmem set to 800,000,000
Swap disabled; swappiness disabled; vm.dirty_background_ratio = 2,
vm.dirty_ratio=5
CPU Governor overridden to “userspace” and forced to run at 4.67GHz on
all 8 cores all the time.
“netstat -s” and “ethtool -S” confirms *not a single packed is dropped
or overrun or missed.*
I am having difficulties finding where is the bottleneck…and why is
not consistent…why it hiccups in 10 minutes or runs great for many
hours….if that is what the error suggests. Benchmarking the RAID 6
shows 1.6GB/s write and 600MB/s read. Also, what is the difference
between *Overflow (overflow) and Overflow (out of sequence )* ? I am
not getting the Overflow (overflow).
I am leading to believe that the RAID might be the culprit...but at
645MB of incoming data, it should even sweat ….
Any advice would be greatly appreciated! Thanks!
Could you perhaps show us the error log?
UHD normally issues "O" for overrun, and "S" for a sequence error.
A sequence error usually means that some part of the network stack
delivered frames out-of-sequence, like sequence 3 being delivered before
sequence 2, etc. Are you
directly connected to your X310, or is there a 10GiG switch in
between your host and the X310?
_______________________________________________
USRP-users mailing list -- usrp-users@lists.ettus.com
To unsubscribe send an email to usrp-users-le...@lists.ettus.com