Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[tickmike]

Whats gone off on this graph at about 19.30 ?.
Again at 1.20 on graph two.

[roseway]

The graphs aren't displaying correctly, so it's a bug. If you change to a shorter time per page, does it revert to a normal display?

[tickmike]

The graphs aren't displaying correctly, so it's a bug. If you change to a shorter time per page, does it revert to a normal display?

No.

seems to be recorded like that, graphs as like that in the file folder as well.

[roseway]

I can only suggest that you save any data which you want to keep, and restart DSLstats. Without some reproducible sequence of events leading to the issue, finding the cause would be a needle in a haystack job.

[kitz]

I recall seeing something similar to this - but not as bad- on one of tickmikes earlier graphs from about a week or so ago. 
I just assumed it may be something to do with the monitoring time/frequency.  ie more often than every min?

[burakkucat]

. . . but not as bad- on one of tickmikes earlier graphs from about a week or so ago. 

Yes, I also saw that graph. My thoughts then were along the lines of "incorrect timekeeping/time dilation/time extension".

Now, I'm not so sure. 

[kitz]

Ive just played with my version, but no matter how I squeeze/extend/resize the graph.... or change the time per page,  I cant get it to do anything like that.   
The only thing I havent tried changing is the sample rate -  I keep that at 60 seconds and never touch it,  because I dont want to cause a conflict with HG612 modemstats.

---
ETA

I note the top graph is showing as having been set at show 1 hour per page.   Whereas the bottom 2 must be set at 12 hours per page.   
Wonder if somehow it has got confused because mike has changing the time per page.     However saying that..  you would think it would have just done that for me when I tried changing mine. 

[burakkucat]

The only way I could emulate the visual aspect of the graph would be to take the raw data and then randomise some of the entries. Re-plotting the data would then produce a similar effect.

[roseway]

The effect appears to be the result of time periodically going backwards. The thin lines are infills which occur when there are some samples missing. Looking at the lowest part of the first graph, there's a green line going down from 19:27 to ~19:40, then it seems to jump backwards to 19:27 again (with an infill between the two points) after which it climbs again to ~19:40. This could be the result if the samples were rearranged in the array which holds them, but how that could happen I can't imagine.

[krypton]

Time jumping forward happens during a connection loss/resync if the sampling period is below 60 seconds. Under this condition dslstats adds 1 minute at each sample and jumps back to the actual time once the connection gets back.

[tickmike]

I note the top graph is showing as having been set at show 1 hour per page.   Whereas the bottom 2 must be set at 12 hours per page.   
Wonder if somehow it has got confused because mike has changing the time per page.     .

I have 12hrs per page and sample 10sec.(now changed to 30sec ).
Yes I expanded it for this post to try to see it better.

[kitz]

sample 10sec.

I wonder if that had something to do with it.   
I know on my PC which has a fairly fast CPU, it can often take circa 5 seconds.   If its going every 10 secs it may not have quite finished one before starting the other.   Eric is probably best to answer...  but I dont know what would happen in those circumstances..  especially if the PC was busy, would the results start to back up?

[burakkucat]

. . .  sample 10sec.(now changed to 30sec ).

That is the problem. Set the sampling rate to be no faster than once per minute and all will be well again.

[roseway]

That's probably a correct diagnosis, but 30 seconds should be fine. With a good connection between the computer and the modem the main time-consuming part of sample collection is logging in to the modem; even a low powered device like a Raspberry Pi will process the data in a short time. If the connection between the computer and the modem is wireless, then all bets are off.