Re: [Observium] error graph related question
any suggestion?
On Sun, Jan 22, 2017 at 3:03 PM, Satish Patel satish.txt@gmail.com wrote:
In Interface Error and Discard graph what is stand for "m" i have seen value like 100m so what does that mean, i can understand "1k" mean 1000 but does "m" means million?
Find attach graph in email.
I've been meaning to ask the same thing, for two reasons. One is just due to typical abbreviations, such as "m" for milli, and "M" for mega, the other is because for us the error graphs appear way off, even though other graphs are not. We might have 100,000 absolute errors on a port, but the error graph will read about "15m". I find the difference in the absolute versus graphed numbers confusing. Perhaps it's an averaged amount over the sample period? If so, can we get actual numbers?
Part of the problem might be the only ports we have with errors are Brocade ports; our Cisco ports don't typically accumulate errors. Perhaps the Brocade error counters are atypical, even though "good" traffic counters are correct.
On 01/23/17 08:11 PM, Satish Patel wrote:
any suggestion?
On Sun, Jan 22, 2017 at 3:03 PM, Satish Patel satish.txt@gmail.com wrote:
In Interface Error and Discard graph what is stand for "m" i have seen value like 100m so what does that mean, i can understand "1k" mean 1000 but does "m" means million?
On Mon, Jan 23, 2017 at 09:07:28PM -0800, William Bauer wrote:
I've been meaning to ask the same thing, for two reasons. One is just due to typical abbreviations, such as "m" for milli, and "M" for mega, the other is because for us the error graphs appear way off, even though other graphs are not. We might have 100,000 absolute errors on a port, but the error graph will read about "15m". I find the difference in the absolute versus graphed numbers confusing. Perhaps it's an averaged amount over the sample period? If so, can we get actual numbers?
Part of the problem might be the only ports we have with errors are Brocade ports; our Cisco ports don't typically accumulate errors. Perhaps the Brocade error counters are atypical, even though "good" traffic counters are correct.
The rates are always in terms of seconds (average over the polling period). So when something like an error is less frequent than once per second, the rate is a decimal per second. So the reciprocal will be average seconds per error which is easier to think of in this situation, or multiplying by 300 to put it in terms of the sample period will also help. -Nick
Nick,
Does that mean "m" stand for minute right? Or million?
Sent from my iPhone
On Jan 24, 2017, at 12:56 AM, Nick Schmalenberger nick@schmalenberger.us wrote:
On Mon, Jan 23, 2017 at 09:07:28PM -0800, William Bauer wrote: I've been meaning to ask the same thing, for two reasons. One is just due to typical abbreviations, such as "m" for milli, and "M" for mega, the other is because for us the error graphs appear way off, even though other graphs are not. We might have 100,000 absolute errors on a port, but the error graph will read about "15m". I find the difference in the absolute versus graphed numbers confusing. Perhaps it's an averaged amount over the sample period? If so, can we get actual numbers?
Part of the problem might be the only ports we have with errors are Brocade ports; our Cisco ports don't typically accumulate errors. Perhaps the Brocade error counters are atypical, even though "good" traffic counters are correct.
The rates are always in terms of seconds (average over the polling period). So when something like an error is less frequent than once per second, the rate is a decimal per second. So the reciprocal will be average seconds per error which is easier to think of in this situation, or multiplying by 300 to put it in terms of the sample period will also help. -Nick _______________________________________________ observium mailing list observium@observium.org http://postman.memetic.org/cgi-bin/mailman/listinfo/observium
The graph scale is calculated by rrdtool. It closely follows the SI/metrix scale
yotta Y zetta Z exa E peta P tera T giga G mega M kilo k hecto h (I've not seen rrdtool display this) deca da (I've not seen rrdtool display this) deci d (this gets no scale as it is a "pure unit") centi c (I've not seen rrdtool display this) milli m micro ยต nano n pico p femto f atto a zepto z yocto y
The 'm' stands for milli. In something like an interface error graph, with an occassional error encountered, you might get a very low number/scale on the graph. 1 errored packet in a 5 minute polling window becomes 1/300 = 0.0033. rrdtool would choose to draw this and automatically adjust the scale so that it would read 0.33 millipackets or maybe even 330 micropackets.
Hope this helps.
Michael
On 25 Jan 2017, at 12:56 am, Satish Patel satish.txt@gmail.com wrote:
Nick,
Does that mean "m" stand for minute right? Or million?
Sent from my iPhone
On Jan 24, 2017, at 12:56 AM, Nick Schmalenberger nick@schmalenberger.us wrote:
On Mon, Jan 23, 2017 at 09:07:28PM -0800, William Bauer wrote: I've been meaning to ask the same thing, for two reasons. One is just due to typical abbreviations, such as "m" for milli, and "M" for mega, the other is because for us the error graphs appear way off, even though other graphs are not. We might have 100,000 absolute errors on a port, but the error graph will read about "15m". I find the difference in the absolute versus graphed numbers confusing. Perhaps it's an averaged amount over the sample period? If so, can we get actual numbers?
Part of the problem might be the only ports we have with errors are Brocade ports; our Cisco ports don't typically accumulate errors. Perhaps the Brocade error counters are atypical, even though "good" traffic counters are correct.
The rates are always in terms of seconds (average over the polling period). So when something like an error is less frequent than once per second, the rate is a decimal per second. So the reciprocal will be average seconds per error which is easier to think of in this situation, or multiplying by 300 to put it in terms of the sample period will also help. -Nick _______________________________________________ observium mailing list observium@observium.org http://postman.memetic.org/cgi-bin/mailman/listinfo/observium
observium mailing list observium@observium.org http://postman.memetic.org/cgi-bin/mailman/listinfo/observium
On Tue, Jan 24, 2017 at 08:56:15AM -0500, Satish Patel wrote:
Nick,
Does that mean "m" stand for minute right? Or million?
Sent from my iPhone
It stands for milli, so 0.001/s. So 1m would be one per thousand seconds average. The Y axis is the same as ever even for events that are less often than once per second. https://www.nist.gov/pml/weights-and-measures/prefixes
LLDP, CDP, and STP are great examples of what can cause port errors less often than once per second in a very consistent way. I had a new circuit turned up that I was running VLAN tagging on, and the providers router was sending me LLDP which was naturally untagged and counted as an error by my router.
With the LLDP timer sending an error to me every 30 seconds, it was a perfectly flat red line at .0333... errors/s. Luckily I could sniff the traffic in JUNOS and see what it was and get the provider to turn it off! :)
-Nick
participants (4)
-
Michael -
Nick Schmalenberger -
Satish Patel -
William Bauer