- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page
Re: Bonding/VLAN bug in 6.6.0
- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Bonding/VLAN bug in 6.6.0
I have found a UI bug (IMO anyway) with the Network/Bonding/VLAN setup in 6.6.0.
First the disclaimer (caveat): This is running on a modified, unsupported (ancient) Pro Pioneer, upgraded with a 3.0 GHz E7600, 2GBytes 1333MHz DDR2, VGA connector, fan mods., etc. The only non-factory build APP installed is linux-dash. None of this has anything to do with the bug, but I don't want to be accused of leaving anything potentially important out, so... Note: Full details below, FYI only.
Attached is a pic. that should say it all! While playing around with various bonding modes (lots of reboots with a single NIC attached to recover admin control), I now have lingering "bonds" with a single NIC (eth0 or eth1 ONLY). I can't delete them, change the settings, add a VLAN to them, or in any way modify them (just want them to go away!) Anything I try results in a crash dialog popup. The correctly bonded "bond0" works fine, as do VLANs added to it with higher IDs than the stuck 1-NIC "bonds".
BTW: If anyone is curious, the E7600 is pretty fast! Even running maxed-out at a fairly well sustained 110MBytes/sec. I never see CPU throughput hit over 50%. I'm pretty sure the old Seagate 1TB ES drives (7200 RPM, built like tanks!) aren't the bottle neck either. In my case it's clear (again IMO) the NICs are the limiting element. As has been explained before, due to the way bonding/teaming is implemented (even with LACP, 802.3ad thru 2+3 mode), it's hard to see increased net throughput via bonding, even with static LAGs, VLANs, a fast external network (4 teamed, i350 server NICs on a Windows Server 2016 TP5 server box, Z170 chipset, 6x 4TB WD Red/5200 RPM "NAS" SATA-III, 6Gbit, overclocked i3-6300), and a fast Layer 2+ smart switch ( NG M4100-D12G). It is especially diappointing that LACP mode 3+4 (supposed to load balance on xmit from a single client) doesn't appear to work in 6.6.0. I see the following in dmesg.log:
bond0: Setting xmit hash policy to layer3+4 (1)
bond0: option primary: mode dependency failed, not supported in mode 802.3ad(4)
Kernel log FYI on E7600:
Oct 07 12:08:31 SIMATH_NAS kernel: smpboot: CPU0: Intel(R) Core(TM)2 Duo CPU E7600 @ 3.06GHz (fam: 06, model: 17, stepping: 0a)
Oct 07 12:08:31 SIMATH_NAS kernel: Performance Events: PEBS fmt0+, 4-deep LBR, Core2 events, Intel PMU driver.
Oct 07 12:08:31 SIMATH_NAS kernel: ... version: 2
Oct 07 12:08:31 SIMATH_NAS kernel: ... bit width: 40
Oct 07 12:08:31 SIMATH_NAS kernel: ... generic registers: 2
Oct 07 12:08:31 SIMATH_NAS kernel: ... value mask: 000000ffffffffff
Oct 07 12:08:31 SIMATH_NAS kernel: ... max period: 000000007fffffff
Oct 07 12:08:31 SIMATH_NAS kernel: ... fixed-purpose events: 3
Oct 07 12:08:31 SIMATH_NAS kernel: ... event mask: 0000000700000003
Oct 07 12:08:31 SIMATH_NAS kernel: x86: Booting SMP configuration:
Oct 07 12:08:31 SIMATH_NAS kernel: .... node #0, CPUs: #1
Oct 07 12:08:31 SIMATH_NAS kernel: x86: Booted up 1 node, 2 CPUs
Oct 07 12:08:31 SIMATH_NAS kernel: smpboot: Total of 2 processors activated (12235.88 BogoMIPS)
- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Re: Bonding/VLAN bug in 6.6.0
Guess my post was too long to actually read and comprehend??? First, I stated that both bonding and VLANs do WORK, but there is a persistent non-valid batch of false "bonds" with only 1 NIC (which clearlly isn't a "bond" and should never happen). What in the world would my browser cache have to do with this??? Do you think my browser has suddenly developed so much AI that it can autonomously generate kernel CRASH popups when I click on active elements of the GUI??? This is a software bug in 6.6.0....
- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Re: Bonding/VLAN bug in 6.6.0
I agree that the stale bonds that can't be removed looks like a bug in 6.6.0
@simath wrote:
...It is especially diappointing that LACP mode 3+4 (supposed to load balance on xmit from a single client) doesn't appear to work in 6.6.0. I see the following in dmesg.log:
bond0: Setting xmit hash policy to layer3+4 (1)
bond0: option primary: mode dependency failed, not supported in mode 802.3ad(4)
It would load balance on xmit to a single client, not from a single client.
This mode has never been fully LACP compliant, since it allows > 1 gbit of traffic to be sent to a 1 gbit client. Another issue is that fragmented packets can stripe both interfaces, which creates issues on reassembly. So I suspect Debian 8 / Jesse removed support from its drivers.
- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Re: Bonding/VLAN bug in 6.6.0
Sage,
Thanks for your knowledgeable reply. I guess I'll just wait until the next firmware update and see if those false bonds go away. In any case I know exactly what steps to take to cause them to happen, if the NG s/w team is interested.
I was all set to upgrade to another ReadyNAS system, but started looking at the costs (this is for a small business). For significantly less $$$ (compared to a new ReadyNAS 5-series) I was able to put together a couple of i3 servers (running Windows Server 2016) where NIC teaming works (it doesn't yet on Win10). I've fully tested it and am able to get a sustained max. transfer rate of nearly 4Gbps. That means the old Ultra6/Pioneer is a bit of a bottleneck, but I'll simply use it in a purely backup role, rather than as a file server as I'd planned. As has often been stated on this forum, teaming/bonding is really more for reliability, rather than speed, and thus for most users not really useful.
I could have opted to go the 10G route, but unfortunately EVERYTHING related to 10G networking is still way to pricey for me.
- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Re: Bonding/VLAN bug in 6.6.0
I've seen a similar issue in a stock ReadyNAS.
What happens when you try to delete the VLANS?
14002089999?
- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Re: Bonding/VLAN bug in 6.6.0
Hi folks,
I`ll get the same error. After I click on delete to remove VLAN on my RN214 I get a cryptic long error number. Jut like JasonQ13. After that nothing changed.
Best regards
Kalle
- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Re: Bonding/VLAN bug in 6.6.0
Hi Kalle13,
Would it be okay if you can also post a screen capture of the error message. The community might have more inputs regarding your concern.
Regards,
- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Re: Bonding/VLAN bug in 6.6.0
Hi FramerV,
sure, no problem.
I hope you see everything.
Best
regards
Kalle
- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Re: Bonding/VLAN bug in 6.6.0
Hi Kalle13,
I am yet to try this out myself. Our LAB is being transferred as we speak, I will be trying to replicate this as soon as I can.
You could also try checking with NETGEAR support but I think you would be needing at least a per incident support for phone assistance.
Regards,
- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Re: Bonding/VLAN bug in 6.6.0
Hi Kalle 13,
I reproduced the issue in local,
If create VLAN on bond, I saw the issue after reboot unit.
Issue only happen on bonding mode.
and will report to developer.
- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Re: Bonding/VLAN bug in 6.6.0
- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Re: Bonding/VLAN bug in 6.6.0
We are seeing the same problem on two new 3138s. Thank you for the update evan2, do you know when the 6.7.0 build is likely to be released?
- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Re: Bonding/VLAN bug in 6.6.0
Thanks evan2. When I get a chance to look at them I'll try the Beta firmware and post my results. Regards.
- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Re: Bonding/VLAN bug in 6.6.0
Applied the 6.7.0 beta and that fixed the issue on both NAS. Thanks for the heads-up.