v1.4.1 Bug Reports and Comments

[colinhowlin]

Hi

Experiencing high CPU usage on new switch running 1.4.1 from RX thread.

I've disabled STP and Discovery on the switch - as per this thread - but problem persists.

Switch is powering a mix of Cambium ePMP & PMP450 as well as some Ubiquiti gear.
Backhaul is an SIAE AlfoPlus also powered by the switch.

Doesn't seem to impact performance, doing pretty much the same throughput as before switch went in but I hate seeing anything use high CPU cycles!

[sirhc]

@colinhowlin

STP is not the cause so disabling this is not needed but it can not hurt to disable it for now.

Disabling the Discovery Tab on the Device/Configuration Tab would be the suspected cause of the high CPU but you have to tell the switch to reboot after disabling it and let us know if the CPU utilization drops.

Could you tell the switch to reboot after disabling the Discovery Tab and report if your CPU drops to normal?

Also could you describe your network, is this a large flat network segment?

How many devices are in this segment?

Would you be willing to allow Eric to Team View into your computer and look at this switch?

[colinhowlin]

Ok, I'll need to wait until later to reboot - this is a busy site for us.
Once I do that I'll report back.

It's not a flat network segment - it's routed.
It is a busy node however, currently about 450 subscribers on this site.
In total we have 14 AP's on this tower and one backhaul radio - all powered by the Netonix.

We're using a WS-24-400A.
Prior to this we were using 2 WS-12-250A's running 1.3.9 with no CPU issues on either switch.

I would prefer to wait to see if disabling discovery resolves this issue.
If not, I could probably put a public IP on the switch and give you access that way?

[sirhc]

@colinhowlin

Do you have v1.4.1 running on other sites OK or is this the only one?

We look forward to hearing back from you once you reboot the switch with Discovery Disabled.

Please think of a time that would be best for Eric to look at the switch that would be the least impact on your business.

During this time we need to be able to reach you immediately if something would go wrong. I am not saying anything will but I always like to plan for any event that could occur and be prepared to deal with it.

How long does it take to get to this site/switch if need be?

Could you please post up screen grabs of your Configuration Tabs.

[colinhowlin]

I'll be rebooting early tomorrow morning GMT, so about 14 hours from now.

This is the only switch running 1.4.1 - all others running 1.3.9.

I'll let you know how the reboot goes.

[sirhc]

@colinhowlin


Please edit your post and put up your Device/Configuration Tab and Device/Status Tab
So you picked your biggest baddest tower to try a new firmware on

Well if you reboot in the morning that means we will have to wait till then and the time to see.

For us to play your switch/network would be best after your prime time which for my WISP ends at 11PM +/-.

I think you are 4 to 5 hours ahead of us so that would be 7-8PM our time?

Or we could do this on the weekend?

[colinhowlin]

Config and Status tabs below.
It's not our biggest site - it's our second biggest!

I think your timezone is EDT?
That would put you 5 hours behind.
2pm for us would be 9am for you - this would be an ok time for us if that suited.

Tomorrow won't suit as I'll be busy with other projects so won't be able to get to site quickly if needed.
Friday should be ok however if you are available.

[colinhowlin]

Here's the output of top - you can see RX Thread eating up a fair chunk of CPU time.

If you need any further info let me know.

[colinhowlin]

After a reboot, CPU usage has returned to more normal levels, averaging around the 30% mark at the minute.
So it would seem Discovery tab is the issue?

[sirhc]

After a reboot, CPU usage has returned to more normal levels, averaging around the 30% mark at the minute.
So it would seem Discovery tab is the issue?

I posted v1.4.2rc4 which now should properly allow you to enable and disable the Discovery Tab.

Once you install this we will need to arrange access to this switch during an OFF PEAK time but also a time that if something were to go wrong you would be free to get to the tower quickly if needed?