v1.4.0rc1-16 Bug reports and comments

[IntL-Daniel]

WS-12-250-AC - v1.4.0RC10: Throughput graph is broken.

See the picture:
* this is total throughput of all ports...there must be the same up and down (at the moment of the screenshot is OK)
* looks like there are no communication breaks but in fact there are not!
* values in the graph correspond to the values in the up right corner, so there are probably bad readings values somehow?
* we are rolling back to 1.3.9 this night (I hope it will go without an issue from 1.4.0RC10?)

[sirhc]

I have v1.4.0rc10 on many of my WISP switches at my WISP although most are WS-24-400A and do not have that issue.

However I do have a WS-12-250-AC at RF Armor warehouse and do notice the total graph throughput you are referring too.

I do not think it is anything to be alarmed about and should not affect performance but is "possibly" a bug that affects the UI graph.

I will talk to Eric about it and see if there is indeed a bug in the UI stats so we can get rc11 out if need be.

I would not roll back I would wait and roll forward but there is no issue with downgrading to v1.3.9 as we tested that several times.

Good "possible" catch Daniel, I will investigate further.

[lligetfa]

WS-12-250-AC - v1.4.0RC10: Throughput graph is broken.

See the picture:
* this is total throughput of all ports...there must be the same up and down

I don't understand why you think they should be the same. They sometimes are the same but not always.

You say the screenshot shows it is OK, so why post it and not one that shows the problem? It may helps us to understand what you are trying to say.

I don't know what Chris is seeing and I don't.

[sirhc]

Your correct Les that it is possible to be not the same but would require a lot of Ethernet errors or drops from a massive mis-configuration or a lot of something?

Would be nice if he showed his full Stats Tab

[sirhc]

I will downgrade the RF Armor switch to v1.3.9 and see.

But my bigger towers running WS-24-400A switches v1.4.0rc10 Total Throughput graphs are the same but it is a different type of traffic than on the RF Armor WS-12-250-AC

[lligetfa]

Not all drops are errors per se. If he has storm control enabled that could show a discrepancy on the graph. Mind you , I don't know what the error counters would show.

[wisp-wand]

wayne

thanks, i am investigating this..for sure the latest rc code has vast improvements, but it seems i have to track down some floating ghost of a bug....

dave

Hi Dave, thanks, I had a couple more, the date and time is wrong as its looking for the production network NTP server which is can't find from the bench!
Jan 1 05:04:07 switch[794]: 48V is out of range (45.9V)
Jan 1 05:05:00 switch[794]: 48V is out of range (32.6V)

Again the switch doesn't react, is that part of the enhancement in the new version, that the spurious values are logged, but the switch no longer knee jerk reactions and starting shutting stuff down?

Wayne

[sirhc]

Not all drops are errors per se. If he has storm control enabled that could show a discrepancy on the graph. Mind you , I don't know what the error counters would show.

True,, many reasons the Total Graphs could be off between Tx and Rx such as the MTU is too small and a lot of packets get dropped or bandwidth limiting which causes packets to get dropped, or he has flow control turned off and a lot of packets are being dropped.

MANY REASONS COULD CAUSE THIS.

TURN ON FLOW CONTROL PEOPLE!!!

[sirhc]

ehh, downgraded to v1.3.9 and now I am upgrading to v1.4.0rc10 again to compare.

As there is not a lot of traffic there on Sunday it is hard to tell.

[sirhc]

Or if he does not have all the ports selected to stats on the Ports Tab then packets would get counted on say the Rx on port 2 and skip the packets on Tx on the other port.

He needs to post his Tabs:
Status Tab to see possible errors
Port Tab to see Stats counted and Bandwidth settings
Device Config Tab to see Storm Control settings

Multicast and IGMP traffic could also skew these number.

I suppose the Discovery option could cause this too?