Solved Troubleshooting poor network performance
-
Below is a screen capture of my NFS repos on Xen Orchestra, and the respective performance that I'm getting according to XO. I'm trying to sort out where my performance issues are stemming from and looking for some guidance.
All remotes were just re-tested for consistency.
After some digging around, I looked at my ESXi host (which is where XO is being run from - so its outside of the pool), and I noticed a 10MB Full Duplex NIC within ESXi.
I've looked at the switch port (an Aruba 2530) and find Port 34 at 10FDx.
If I attempt to manually set the port speed to 1000 FD, I get "Value 1000-full is not applicable to port 34"
If I check the network interface within XO, I get the below, which is not 10FDx...
A CLI Speed test shows
Would the next assumption be that this port is connected to a HUB and that physical re-cabling is going to be needed to address this?
-
@DustinB3403 said in Troubleshooting poor network performance:
Moving from that port which was only giving 10FDx to an unused port, gave us 1000FDx.
I'm not sure where this issue stems from..
Got it sorted out, for some reason (and I'm still working on the specifics) our ESXi hosts secondary NIC keeps falling to 10FDx (likely some misconfiguration at setup).
I've moved XO off of this nic, and performance has been fixed.
-
In looking at the interface further, I do have quite a lot of dropped packets.
The switch has been up for 224d 16h 48m 21s
-
The Dropped Tx percentage is 0.6 of all packets dropped...
-
@DustinB3403 I would think a hub would show half-duplex not full duplex?
-
@jt1001001 I'm in office today, and moved the port from 34 to 33 and am now getting 1000FDx on the port and within ESXi.
So it's very odd.
-
Moving from that port which was only giving 10FDx to an unused port, gave us 1000FDx.
I'm not sure where this issue stems from..
-
@DustinB3403 said in Troubleshooting poor network performance:
Moving from that port which was only giving 10FDx to an unused port, gave us 1000FDx.
I'm not sure where this issue stems from..
Got it sorted out, for some reason (and I'm still working on the specifics) our ESXi hosts secondary NIC keeps falling to 10FDx (likely some misconfiguration at setup).
I've moved XO off of this nic, and performance has been fixed.
-
D DustinB3403 has marked this topic as solved on
-
@jt1001001 said in Troubleshooting poor network performance:
@DustinB3403 I would think a hub would show half-duplex not full duplex?
It would definitely. A hub cannot be FDX.
-
@DustinB3403 said in Troubleshooting poor network performance:
@DustinB3403 said in Troubleshooting poor network performance:
Moving from that port which was only giving 10FDx to an unused port, gave us 1000FDx.
I'm not sure where this issue stems from..
Got it sorted out, for some reason (and I'm still working on the specifics) our ESXi hosts secondary NIC keeps falling to 10FDx (likely some misconfiguration at setup).
I've moved XO off of this nic, and performance has been fixed.
If you want to improve ESXi performance, install KVM.
-
@scottalanmiller said in Troubleshooting poor network performance:
@DustinB3403 said in Troubleshooting poor network performance:
@DustinB3403 said in Troubleshooting poor network performance:
Moving from that port which was only giving 10FDx to an unused port, gave us 1000FDx.
I'm not sure where this issue stems from..
Got it sorted out, for some reason (and I'm still working on the specifics) our ESXi hosts secondary NIC keeps falling to 10FDx (likely some misconfiguration at setup).
I've moved XO off of this nic, and performance has been fixed.
If you want to improve ESXi performance, install KVM.
Yea, that's a different conversation entirely, I do want a outside of the XCP-ng pool environment, in case something goes sideways. I'm dealing with some sunkcost conversations about it, though I am making progress.
