Hi everyone,
I set up a server on older hardware I still had lying around (XEON E3-1225, 16GB DDR3, Intel S1200KP) and the system seemingly runs fine.
Used fio to check the the wild mixture of HDDs works as intended and got about 150MB/s write speeds which do the trick for me.
However I want to use this box as a backup target for my main TrueNAS system so I set up some replication tasks.
Unfortunately they are copying at a speed of about 10MegaBITS which is way to slow for the network speeds I think should achieve.
The network port the system is connected to reports 1Gbe, the system itself shows 1000Mb/s - so this seems fine.
CPU idles at around 1-5% and memory seems fine, too (13GiB free).
If I run iperf3 against that box I get an average speed of 10,6Mbits/sec in both directions which is obviously way to low.
Running iperf3 agains my other trueNAS gives me 980+ MBits/sec which makes me believe that neither my desktop computer nor the other TrueNAS instance have an issue.
yeah if iperf itself is stuck around 10mbit both ways, the replication task isnt the problem. the path to that box is.
old intel boards like the s1200kp sometimes lie a bit in the ui. id check on the slow box:
ethtool
ethtool -i
look for Speed: 1000Mb/s and Duplex: Full for real. if it says 10 or half duplex, swap the cable and try another switch port. cheap patched cables love doing exactly this.
also confirm youre hitting the onboard nic you think you are (some of those boards share a port with bmc / have two jacks). and try:
ethtool --set-eee eee off
once. eee has made a few “reports 1g, transfers like dialup” boxes for me.
if ethtool really shows 1000/full and eee off and a known-good cable still gives ~10mbit iperf, post the ethtool output + which iface / driver. then its probably the nic/driver on that xeon box, not your main truenas.
Here is what ethtool gives me. I think that looks ok?
ethtool enp3s0 Settings for enp3s0: Supported ports: [ TP ] Supported link modes: 10baseT/Half 10baseT/Full 100baseT/Half 100baseT/Full 1000baseT/Full Supported pause frame use: Symmetric Receive-only Supports auto-negotiation: Yes Supported FEC modes: Not reported Advertised link modes: 10baseT/Half 10baseT/Full 100baseT/Half 100baseT/Full 1000baseT/Full Advertised pause frame use: Symmetric Receive-only Advertised auto-negotiation: Yes Advertised FEC modes: Not reported Link partner advertised link modes: 10baseT/Half 10baseT/Full 100baseT/Half 100baseT/Full 1000baseT/Full Link partner advertised pause frame use: No Link partner advertised auto-negotiation: Yes Link partner advertised FEC modes: Not reported Speed: 1000Mb/s Duplex: Full Auto-negotiation: on Port: Twisted Pair PHYAD: 1 Transceiver: internal MDI-X: off (auto) netlink error: Operation not permitted Current message level: 0x00000007 (7) drv probe link Link detected: yes
I do not see anything about EEE. And it looks like this chip does not support EEE:
ethtool --show-eee enp3s0
netlink error: Operation not supported
I have been using the upper network port.
However I have tried the second network port below and that works as intended. I get 935MBits on average which is perfectly fine.
After that I switched the cables around to rule out an issue with the cable from the first tests. The second network interface still tested fine with about 925MBits. The first one, which was the slow one, did not do anything with iperf3 but maybe that is a firewall issue since I am currently testing against a different machine. BUT: the zfs replication is running on full chooch right now. Network utilization shows 984MBits the whole time. Maybe it was just stuck on something strange…
I assume the network disconnect/reconnect solved it but if anybody has a better explanation for it I am open for it
yeah if ethtool really says 1000/full then the cable story is less likely. next stuff id try on that slow box:
ethtool --show-eee enp3s0
ethtool --set-eee enp3s0 eee off
then kill the fancy offloads for a minute and retest iperf:
ethtool -K enp3s0 tso off gso off gro off
also post ethtool -i enp3s0 and whether that jack is the only one (s1200kp boards sometimes have a weird second path). if eee/offloads change nothing and a known-good cable+port still sit at ~10mbit both ways, id start suspecting the nic/driver on that old xeon board more than your main truenas.
