01-08-2019 12:56 AM - edited 01-08-2019 01:01 AM
I could make an argument both ways on this, but I'm in agreement that users should not just automatically upgrade for the sake of upgrading. Yes I have those ER12Ps running on 2.0.0 now, but they aren't in production and I've been testing them in a realistic environment connectrd to the rest of the routers they will be installed with for a while, and plan to continue doing so for the next week or so.
Frankly, anyone who just blindly upgrades production devices on the first day of a release is asking for trouble. Or to loose their job.
Aww come on, sure I agree with what you say but pushing out half baked firmware as stable in order to gather more data is just plain vanilla reckless and I think someone over at UBNT has lost the plot. I have hundreds of these devices deployed and wanted to try this on just one of them, I've already recovered, I just swapped out the unit, but many out there won't be in such a position of luxury and it's not as though I can fire myself. Yes you can put two identical devices on a bench, upgrade them with identical firmware, one will upgrade fine, the other will brick because spacetime faaaarts. The point I'm making here is that firmware that bricks devices on upgrade shouldn't even make it out of alpha. To infer the end user is somehow being reckless for upgrading with STABLE firmware is just absurd.
01-08-2019 02:10 AM
01-08-2019 04:10 AM
From the release notes:
[Offload] - Add offloading support for bonding interfaces for Cavium-based routers with "system offload ipv4 bonding enable" config settings
Is this just for the more recent models (e.g., the ER-4), or are earler units (specifically, the ER-Lite) also included?
Also, what exactly is meant by "bonding"? Link aggregations (e.g., LACP), and/or bridging multiple interfaces?
And no support for this for IPv6?
The bonding link ties the switch chip
in the ER-12’s to the router.
01-08-2019 04:29 AM
The bonding link ties the switch chip
in the ER-12’s to the router.
Actually the bonding offload is indeed for interface aggregation, not bridging.
I believe it was mentioned that this applies to all Cavium-based routers and not just the latest models. This came with the newer version of the Cavium SDK that is now in use.
I'm not sure why it is listed under layer 3 (ipv4) but that is where Cavium placed it:
user@lab-erl3:/proc/cavium$ ls -ld /proc/cavium/ipv*/bonding -rw-r--r-- 1 root root 0 Jan 8 07:24 /proc/cavium/ipv4/bonding -rw-r--r-- 1 root root 0 Jan 8 07:24 /proc/cavium/ipv6/bonding
01-08-2019 04:50 AM
it looks like that 2.0.0 firmware for ER-X / ER-X-SFP got kernel from future .
Upgraded linux kernel to v4.9.79 for Octeon-based routers (ER, ER-pro, ER-lite, ER-PoE, ER-Infinity)
Upgraded linux kernel to v4.14.128 for Mediatek-based routers (ER-X, ER-X-SFP)
Latest 4.14 series is 4.14.91 (at least at the time of writing this reply).
Now I understand why there are bugs and users flag it as a BETA .
01-08-2019 05:07 AM
I'm not sure why it is listed under layer 3 (ipv4) but that is where Cavium placed it:user@lab-erl3:/proc/cavium$ ls -ld /proc/cavium/ipv*/bonding -rw-r--r-- 1 root root 0 Jan 8 07:24 /proc/cavium/ipv4/bonding -rw-r--r-- 1 root root 0 Jan 8 07:24 /proc/cavium/ipv6/bonding
Bonding won't enable unless IPV4 offloading is enabled, could be a reason they placed it there.
01-08-2019 05:32 AM
What's more problematic is that there has been zero response from any UBNT staff. The least we should expect when problems arise from firmware they push out as stable is that they are listening and responding.
01-08-2019 05:34 AM
That's what i mean. At work i upgrade firmwares only after they are a few months out in the wild.
But if i brick my home device..i'm lost (i have only 1 device). That's why i do the same in my house.
And a lot of people doesn't have the redundancy to test. They see an official 2.0 release, update... and problems!
01-08-2019 05:42 AM
I decided to test this on one of my many ER-X devices. It came back up (connected devices are contactable) but the ER-X itself now has no GUI or SSH access. It's doesn't even answer pings. Tried a power cycle by resetting the upstream port PoE, but still no go.
FYI I have this device on my desk now. A factory reset brings the GUI back to life and it's on version 2.0.0 according to the GUI. However, setting up via the switch wizard and then rebooting makes the GUI and SSH fail again. I'll reset once more and downgrade.
01-08-2019 06:07 AM - edited 01-08-2019 06:07 AM
Further to the above, this device came back to life with a reset, downgrade to 1.10.8 and restoration from a backup. No need to get the screwdrivers out. Now redeployed.
01-08-2019 06:10 AM
Upgraded my ER-X from 1.10.8 to 2.0.0 via the EdgeOS GUI. Up and running without issues for almost 24 hours. My setup is very basic: IPv4 only, DHCP off, no VPN, no special QoS rules.
01-08-2019 06:36 AM - edited 01-08-2019 06:45 AM
Although I've seen no complaints from Edgerouter 4 users, I am still hesitant to put this firmware on, given 4 pages of issues and it's only been out under 24 hours. Seems like the majority of issues are from ER-X users however.....
AP AC LITE
UAP nanoHD (x2)
01-08-2019 06:38 AM - edited 01-08-2019 06:39 AM
Indeed, I planning to test on some other devices in my network. Now I've been sucked into this beta test, I feel compelled to continue :-)
01-08-2019 06:49 AM - edited 01-08-2019 07:02 AM
One of the users on this thread had issues with an ER4 that made him go back to 1.10.7...I have an ER 6p and won't be updating to this firmware. I'm on 1.10.8 and it's working just fine. I'm wondering if Ubiquiti will pull this firmware given the amount of issues described here.
01-08-2019 06:55 AM
Upgraded my home ER-X via cli, deleting system image prior. No issues, however just like the later betas DHCPv6-PD seems broken, at least on the ERX line. The last time I had working IPv6 was on alpha.2. No issues on 1.10.8
This the only thing in my dhcpv6-pd log on 2.0.
Jan/08/2019 06:42:37: dhcp6_ctl_init: bind(control sock): Address already in use Jan/08/2019 06:42:37: client6_init: failed to initialize control channel
Rolled back to 1.10.8 and IPv6 working again.
01-08-2019 07:08 AM
Although I defended 2.0 beta 3 acting more like an alpha release, this stable 2.0 release is troubling... Beta 3 is far from stable, and one-two weeks is certainly not enough time to fix all issues that were raised in beta 3.
Now that we are encouraged to keep devices updated to address security concerns, I'm sure that plenty of users will have unpleasant surprises, and this may hurt ubiquiti's reputation.
Comments about the ancient kernel; I agree, however the product functioned as intended and was more secure than a barely tested product...
As for comments regarding having no job is it were'nt for bugs; fixing stuff should be an unintended opportunity, not an expectation from products being crap.
It often happens that X.0.0 releases will have bugs, however those bugs shouldn't be known at GA release, they should simply be found through extensive use and should not affect core/basic functionality (as this should be thoroughly tested).
I'm not able to see the stable 2.0 release in the downloads list... so unless I missed something, at least there's that!
01-08-2019 07:12 AM
> ERX ... on switch0 one of the vlans on the ports (that I think was configured before the update) was missing afterwards so the connected device wasn't accessable.
Are you sure that configuration was saved? If you configure something from CLI then you need to run `save` explicitly otherwise config will be lost after reboot.
> Device on eth2 couldn't obtain DHCP lease
I was able to reproduce similar issue - if `vlab-aware` is enabled then DHCP responce fails on non-vlan switch ports. Fix will be available in v2.0.1
> Any eta for more than 8 wans on offload engine?
I suppose you are talking about LoadBalancing instead of offload? We are working to remove 8-interface limit in LoadBalancing and make it dynamic. It is targeted to v2.0.2
> I can confirm that dhcpv6-pd is not working correctly.
There have been report on beta forum about radvd crash that caused problems with radvd, however we could not reproduce this issue in lab environment we are still investigating it.
Could you please check what if you see also see radvd crash messge in /var/log/messages?
> Why is this being released with so many serious known issues? Suggest to pull this back to beta.
> As is, the release will receive negative press, confusion and frustration.
Upgrade from v1.10.x to v2.0.x is major change because affects all areas in ER (kernel, routing, offload, firewall, bootstrap), and I acknowledge that there were some known issues affecting certain functionalities mostly on ER-X devices.
But we need v2.0.0 to be available for new EdgeRouter models that won't otherwise work with v1.10.x firmware.
> ER-X ... Dashboard tab of the GUI it now shows 0bps constantly for eth1, eth2, eth3 and eth4
Does eth1, eth2, eth3 and eth4 belong to switch0?
> What are the differences to the last beta version?
OSPF - Fixed regression in v2.0.0-alpha.1 that caused OSPF failure if OSPFv3 was configured
RIP - Fixed regression in v2.0.0-alpha.1 that caused RIP failure if RIPNG was configured
IGMPProxy - Fixed regression in v2.0.0-beta.3 that caused IGMPProxy not to start even if it was configured
RSVP - Fixed regression v2.0.0-alpha.1 that caused RSVP not to start
> I am remotely managing a ERX with a UAP-AC-Lite connected in passthru mode on eth4
> it is not pingable from the CLI
- Who was not pingable ER-X or UAP-AC-Lite?
- Did non-pingable device had LED light blinking?
> performance degradation with all of the 2.0 software versions.
> v1.10.8 - 914Mbps down, 902 up (maxes out the WAN connection).
> v2.0.0 betas & release - 426Mbps down, 151Mbps up.
I now can reproduce similar symptoms - when I first run iperf3 test it reports 900Mbps, but after running it few more times max bandwidth drops to 600Mpbs. After runnign again and again I see that max thoughput periodically jumps from 900-> 600 -> 900 and so on. We are investigating this.
> This new release broke my real dmz setup please fix rolling back to 1.10.8.
Which model are you talkinga bout ER-X/ER-X-SFP?
> Moving forward, will there be updates/maintenance releases for v1.x.x?
Yes, we will continue supporting v1.10.x so some time
> ER-X itself now has no GUI or SSH access. It's doesn't even answer pings.
> connected devices are contactable
1) Does it mean that ER-X is routing traffic between LAN and Internet?
> It came back up but the ER-X itself now has no GUI or SSH access
> Tried a power cycle by resetting the upstream port PoE, but still no go.
Try login router with ssh-recovery as described here -> https://help.ubnt.com/hc/en-us/articles/360002231073-EdgeRouter-How-to-Use-SSH-Recovery-
> failed! ERL upgrading from beta 3.
> gui reports upload failed, not enough space.
> tmpfs 245768 86212 159556 35% /tmp
There's at least 160Mb of free space in /tmp/ and it's more than enough to make an upgrade.
Please try upgrading from CLI post then post all upgarde messages visible in CLI
> Lighthttpd will not come up when upgrading to 2.0
> 2019-01-08 01:01:34: (plugin.c.229) dlopen() failed for: /usr/lib/lighttpd/mod_websocket.so /usr/lib/lighttpd/mod_websocket.so: cannot open shared object file: No such file or directory
The /usr/lib/lighttpd/mod_websocket.so is present in v1.10.8 and it't missing in v2.0.0, but it's not causing WebGUI failure on my v2.0.0 test routers.
- Do you have custom config in service gui?
- Did you modify /etc/lighttpd/lighttpd.conf by hand?
> For some reason IPv6 took about 10 minutes to come up on my connection, but seems to be working fine after that.
Those symptoms look like a very old issue that was present since "v1.9.1" firmware that has not been fixed yet (Discussed here). Are you sure that IPv6 with your config works instantly after reboot with v1.10.8 firmare?
> using the GUI to make changes to an Advance Queue configuration and have it completely wipe out all vif's on the device
> The ER-12 just dies, no output in the serial console. Have to pull power to reboot and use the console to restore a previous config.
We are looking into this.
> IPv6 router advertisements still inop.... radvd continues to crash as before.
We are not able to reproduce this issue lab environemnt therefore I think it's something very special your environmetn that is causing it. @ubnt-Sandis is investigating this and preparing a special radvd build to track this issue.
> Traffic Analysis/Traffic Counters still not working... Never worked in a single Beta version
> OpenVPN certificates generated by any OpenVPN client version after 2.4.4 are not accepted by this firmware
Those are scheduled to v2.0.1
> Add offloading support for bonding interfaces for Cavium-based routers
> Is this just for the more recent models (e.g., the ER-4), or are earler units (specifically, the ER-Lite) also included?
All cavium-based models support bonding offload
> Also, what exactly is meant by "bonding"? Link aggregations (e.g., LACP)
Yes, it's what you configure with "set interface bonding ..." CLI commands
> And no support for this for IPv6?
Nope, right now no support for IPv6
> After the upgrade on my ER4 2 vpn's to zyxel on the other side are not working.
- Please post you config
- If it's IPSec VPN then please post output of `sudo swanctl --log` when ER fails to connect to Zyxel
> Latest 4.14 series is 4.14.91 (at least at the time of writing this reply).
Sorry, typo. It should be 4.14.54
01-08-2019 07:21 AM
My router is a Netgear R7000 running DD-WRT, and I have 100x10 dsl with Bell Canada. My most recent bandwidth test is within spec of those numbers. To be honest, I don't see much difference with 2.0.0 over 1.10.8 with my setup. I would have thought a major release would maybe spruce up the GUI a bit
01-08-2019 07:32 AM
I have skipped the previous version, that's why the websock seems to be missing.
No, until a few hours ago I had had a standard lighttpd conf.
Can I get the official websock please? I think it's better than my homebuilt.