v1.3.9rc2 through rc14 - BETA/Release Candidate

[sirhc]

We reloaded and then rebooted. The SNMP CPU issue went away so it must have been a bad load. Thanks for the help. The only issue I see at this point is that port 1 has a link light on it with nothing plugged in and no power applied to the port. I have attached an image.

Gilbert

That is not a link light, we use that LED as a power LED as it has no other function.

Port 1 has no POE "Out" option which is what the LED is used for on other ports.

[Eric Stern]

Put on rc8 on a new WS-12-250-DC. I changed the MTU on each port to 9600 from the default MTU size, and upon submit I get a "Error applying configuration!" popup. Clicking OK, I get a "Applying configuration... " countdown like normal. Upon refresh I see the MTUs were properly set to 9600. Not sure if this is new in rc8 or not.

I was able to duplicate this. It happens if you change the MTU on the port that the MAC is connected to. When the MTU is changed the link is dropped briefly and that causes Safari to lose the connection and display the error dialog. As you noticed the configured is saved and applied correctly.

This is a bug in Safari (or at least an unwanted feature), and I can't think of a way to work around it. But I think I'm not going to worry about it since:
- it only happens when using Safari on OSX (Firefox and Chrome on OSX work fine, all browsers on Windows work fine)
- it only happens if the computer is directly connected to the switch
- it only happens if changing the MTU or link speed on the port the computer is connected to
Thats a very specific configuration that won't happen very often. It'll basically only ever happen on the test bench, since if you are remotely managing the switch you won't be directly connected to it.

[bmv]

Upgraded a 8-port AC switch from 1.1.8(?) to v1.3.9rcX and it completely got messed up.

Voltages and current readings reported errors and weren't showing anything.
So we decided to reboot it from the dropdown menu.

It then didn't come back online and needed a manual power cycle after waiting a good 10 mins.

Fortunately someone was nearby at the time.

Once power cycled, it all looks okay.

Is the a specific upgrade path required as we upgraded another switch that was on 1.3.8 without any problems earlier in the day.

[sirhc]

Upgraded a 8-port AC switch from 1.1.8(?) to v1.3.9rcX and it completely got messed up.

Voltages and current readings reported errors and weren't showing anything.
So we decided to reboot it from the dropdown menu.

It then didn't come back online and needed a manual power cycle after waiting a good 10 mins.

Fortunately someone was nearby at the time.

Once power cycled, it all looks okay.

Is the a specific upgrade path required as we upgraded another switch that was on 1.3.8 without any problems earlier in the day.

It is not uncommon for some "small" percentage of firmware upgrades to have an issue and may require a reboot and in some even "much" rarer event require a factory default or even possibly a TFTP recovery . I have seen issues with many manufacturers including Cisco that required a reboot over the years.

My guess is you just got un-lucky.

There are over 7,000 switches out there in service and every so often you hear of someone having an issue with a firmware upgrade. Being a WISP for almost 20 years I have grown used to the fact that if you have 500 customers completely happy maybe 1 or 2 will make an effort to say "good job" but that 1 or 2 customers that has an issue will make it their life mission to tell everyone.

[cbl]

Here's a UI bug (not a show stopper by all means). Running latest RC. I click the 30sec/5min/1hr options down on the graph in order with a port selected.. I go back to 30sec and the interface throughput or packetrate graphs dont refresh back from 1hr options. In fact, I have to reload the page to get the graphs working again. Probably some JS weirdness. Running Safari on OSX. Able to reproduce with Chrome too.

[sirhc]

Here's a UI bug (not a show stopper by all means). Running latest RC. I click the 30sec/5min/1hr options down on the graph in order with a port selected.. I go back to 30sec and the interface throughput or packetrate graphs dont refresh back from 1hr options. In fact, I have to reload the page to get the graphs working again. Probably some JS weirdness. Running Safari on OSX. Able to reproduce with Chrome too.

Must be your browser as it works for me?

[nickwhite]

Here's a UI bug (not a show stopper by all means). Running latest RC. I click the 30sec/5min/1hr options down on the graph in order with a port selected.. I go back to 30sec and the interface throughput or packetrate graphs dont refresh back from 1hr options. In fact, I have to reload the page to get the graphs working again. Probably some JS weirdness. Running Safari on OSX. Able to reproduce with Chrome too.

Similar issue here. Running Chrome 47.0.2526.80 on Windows. No Extensions/addons installed.

Select Port 1, shows 30 seconds. Click 5 min, graphs update ok. Click 1 hour, graph displays. Click back to 30 seconds, graph doesn't update (still shows 1 hr graph). Click a different port, still shows 1hr graph from Port 1. Click off of any ports, it shows the main graph, but it's not updating and the RX/TX throughput in top-right corner is not updating.

EDIT: Tested on same switches with IE 11 and it does the same thing.

EDIT: Tested on same switches with Firefox 41.0.2 and it does the same thing. In case it matters, these switches were newly upgraded and the 1hr graph only has about 5 minutes of data.

[zaltec]

Hi Chris,
with the latest rc6 and rc8 we see CPU running at 100% if SNMP is enabled and device is polled by Dude. SNMP remains unresponsive if that occurs.
A reboot is needed but as soon as Dude queries SNMP the CPU goes back to 100%.

BR,
Giorgio

[cbl]

n case it matters, these switches were newly upgraded and the 1hr graph only has about 5 minutes of data.

I can confirm it happens only on newly powered/upgraded switches that do not have a lot of historical data. From my guess, it seems as if there is less than 1hr of full data then the javascript freaks out. I have other RC8's that are behaving properly if they have a bit of uptime.

[Eric Stern]

Hi Chris,
with the latest rc6 and rc8 we see CPU running at 100% if SNMP is enabled and device is polled by Dude. SNMP remains unresponsive if that occurs.
A reboot is needed but as soon as Dude queries SNMP the CPU goes back to 100%.

BR,
Giorgio

Someone else reported that they fixed this problem by reflashing rc8, so you can try that.