SMB connections unstable

Community Support
1

Hi all,

Does anyone have experience with using SMB connections over zerotier?
I seem to have reliability issues (disconnecting) from time to time.
Could this be related to ISP’s traffic shaping udp traffic?

Thanks in advance!

Tim

2

Hi timmmy,
do you have direct, peer to peer connections between the smb client and server? It should generally work, but network file systems like smb, nfs, afp… all seem sensitive to the latency of not being on the same actual LAN.

3

Hi zt-travis,

Thanks!

The list of peers shows the other node as LEAF with direct link and the correct public IP --> I assume this implies a direct connection?

As the severity of the issue seems to fluctuate quite a lot (as do the transfer speeds), it still looks like a provider bottleneck (intentional or unintentional) to me.

I’m now setting up a third node with a hosting provider to check if I can point it down to one of the two connections/ISP’s. First test show results that are way better then what I get end-to-end.

Is there an easy way to have traffic relayed through the third node to see what that gives?

Regards,

Timmmy

4

That’s right. What’s the latency between those nodes? Are they geographically far apart?

Hmm.
Put a little test smb server on the 3rd node?

5

Hi,

It took some time, but finally found some time to test some things.

First of all, we changed ISP’s at one side (unrelated to this), but that already seemed to improve the throughput.

Good idea to set up a smb server on the 3th node. In absolute throughput (Mbps), everything measures out fine. However in one direction sometimes the connections speeds drops every so often. That pointed me to some ping tests…

All nodes are approx 20-30ms apart, but a ping over the zerotier IP’s show this:
rtt min/avg/max/mdev = 24.578/311.084/1037.331/367.330 ms (ISP1 - cloud)
rtt min/avg/max/mdev = 12.329/43.591/322.791/93.066 ms (ISP2 - cloud)
rtt min/avg/max/mdev = 26.025/28.496/33.722/2.345 ms (ISP1 - ISP2)

So the cloud provider does not seem to be helping in testing as the Jitter completely skews the measurement. We’re running OpenStack there… Any known issues with OpenStack?

What I’ve also noticed is that mainly smb browsing speed seems to be an issue. So might indeed be RTT related.

Anyhow, application we were testing is Veeam backup. Somewhere hidden a menu is a option to optimize for LAN or WAN target, with a default to LAN. I’ve changed this now and in combination with the new ISP it is acceptable now.

Other tips & tricks are still welcome, but I guess we’re more or less at the end of what we can change/test.

Regards,

Timmmy

6

This topic was automatically closed after 14 days. New replies are no longer allowed.