Webstats - odd displays

[#1]

.
Webstats have arrived on schedule (and are fine) however the "total sites" and "total locations" have some weird (long) entries in the Kb in, Kb out, columns.
The "sites" ones seem all the same but the "locations" ones aren't (enclose sample screenshots of each).
The stats are perfectly usable as they are, just flagging this up in case somehow Webstats is collecting loads of not-required data on everyone's accounts which might cause the system to become overloaded in the days ahead.

Regards,
Penny.

[#2]

My webstats show similar totals too for June though zero, as they should be, for May. The raw log entries are normal - kB In and KB Out data columns not present - and the individual webstats entries show zero. It looks like the monthly totals might have been initialised incorrectly at the beginning of the month and that is being carried forward.
The problem might go away when July arrives and the totals are reset.
David

[#3]

I'm seeing the same too but one entry oddly has the correct zero value.
Must be picking up bad data somewhere along the line and not handling it in the correct manner. It's the sort of thing you see when you get a negative number when it's only expecting a positive value, you end up with crazy values.

[#4]

Quote from: spraxyt

My webstats show similar totals too for June though zero, as they should be, for May. The raw log entries are normal

Appreciate you checking that the raw logs are okay : )  However when looking at the Webstats this morning (same type of display as yesterday) it struck me that this problem only surfaced yesterday (when the Webstats for 12th June had been processed), ie hasn't been evident every day this month.  I don't have enough grasp of the technical backdrop to have any idea why that should be.

Quote from: RPMozley

I'm seeing the same too.

I'd guessed it would be an across-the-board sort of thing though like you am puzzled why one of your entries should have had the usual zero value (just went and looked back at the png graphics above in case one of them had zero top or bottom, ie beginning or end of the day's stats, like a GMT/BST anomaly, but no, they're all the same type).

Quote from: RPMozley

Must be picking up bad data somewhere along the line and not handling it in the correct manner.

... that was sort of what I'd figured.  The main thing that eventually decided me to make the post on here, yesterday, was that the percentages (in the little Kb columns) were all 0.00%, like each was an infinitesimally tiny fragment of something much bigger, and the displayed numbers are all very long.  I was really just flagging things up in case there's something in the mix that shouldn't be, that might crash the system in due course if not sorted.

Quote from: RPMozley

It's the sort of thing you see when you get a negative number when it's only expecting a positive value, you end up with crazy values.

... that sounds logical (though I don't have any knowledge or experience of the maths).  That type of reason might be more reassuring though than "bad data" in the system (especially in very large quantities); however I don't understand why either sort of issue should have just presented, when it hasn't previously.  After the Kb in/out columns appeared a few years back (Webstats version upgrade? server upgrade?) they were always shown as zero aside from one month, two or three years back maybe, with tiny content-factors iirc, but nothing that looked like the displays now visible.
The odd-display factors don't affect the usability of the stats here (other factors are much as normal) so I don't see any of this as a "problem", as such. Regardless it might be constructive to know if anything to do with the Webstats run was changed on the morning of Friday 13th June (when the stats for 12th June were generated) in case something fundamental might be amiss, somewhere.

Regards,
Penny.

[#5]

Quote from: RPMozley

I'm seeing the same too but one entry oddly has the correct zero value.

I have one of those, this morning : )  or, at least, the first line of both tables "Total Sites" and "Total Sites by Kb F" each has zero in the kB in / kB out columns.
Interestingly, in the pie chart near the foot of the stats, the "unknown" percentage has reduced to 12%, lowest I've ever seen it, and even only half way through June, more "locations" are shown than for the whole of May (or April, or March).
The "unknown" factor has been stuck at around 30% for a couple of years now (since about May 2012 I think) so 12% looks like a significant improvement : D

[#6]

If this behaviour continues into July then shout up folks and I'll see if I can get somebody to do some digging. Unless I'm misinterpreting what's being said, it's not causing too many problems at present is it?

[#7]

No, no problems with it Bob. Just is odd and the worry that it might be a symptom of something amiss. I'm sure the rest feel the same way, if not please say.

[#8]

Quote from: Penny

Interestingly, in the pie chart near the foot of the stats, the "unknown" percentage has reduced to 12%, lowest I've ever seen it, and even only half way through June, more "locations" are shown than for the whole of May (or April, or March). The "unknown" factor has been stuck at around 30% for a couple of years now (since about May 2012 I think) so 12% looks like a significant improvement : D

Time will tell I guess, but I wonder if that change is associated with improved DNS load balancing - fewer timeouts? I assume there is a heavy load when webstats are being processed.

[#9]

They're processed post-Midnight unless I'm mistaken i.e. not during a period of heavy DNS load. There's still the potential for the two things to be related though I guess.