Print

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

[renluop]

I'm still on 5.6.2 and just noticed that the text pages for "Error Stats over 24 hour period! do not show a commencement time.

Unless my memory is getting worse , I'm sure they did. Is it my memory , or was the info withdrawn for some reason?

[roseway]

I can't remember either, but I took a quick look at the source code, and it certainly should preface each 24-hours-worth of data with the starting time and date. I guess it's academic now, because the average errors function has been completely rewritten since that version, and it now records the data by calendar day, not random 24-hour periods.

[renluop]

Remind me, when the change to calendar day took place, please!

Does the new method mean that, say the first stats after midnight are taken @ 01:30. the per day period would be

number so far/1.5*24?

[roseway]

The change was in the pre-release version 5.7.3, which I announced on the 10th March.

The average values which are displayed are the average error rates per hour for the day in question, so the formula would simply be the number so far divided by 1.5 (in your example).

[renluop]

Sorry to be such a nuisance, but in the attachment are snips from the error text logs taken two hours apart, the later being on right hand side. As the clock starts at midnight, I'd have expected the second reading on the right hand to be broadly the same as that on the left, as both would relate to the same sample time.

Patently I am not getting something right in my mind, but can't think what!

[roseway]

Error rates vary over time, so the averages will change. Between your first block and the second one two hours later, error rates must have increased.

[renluop]

But if #1 l.h were taken at 21:00 and #1 r.h at 23:00, surely #2 r.h. would relate to the same sample as #1 l.h. That is if both sets of figures have midnight + 1 sec as their origins.

[roseway]

I'm sorry, I don't know what you mean. Is it important? As I said, that error averages system is no longer in use.

[renluop]

Sorry for my worrygutting! 5.8 now installed. It does look far simpler for errors.

[renluop]

Just noticed that both error stats and connection stats text are being generated twice in snapshots at each sampling, about 1-2 secs apart. A glitch here, or with DSLStats?

[roseway]

I've enabled those two items on my system to see what happens. I'll be back later.

[roseway]

I can confirm the duplication of the average errors text, and I've found out why and fixed it. I can't confirm the duplication of the connection stats text though.

In looking at this I've realised that there is an inconsistency in the file naming of the average errors text, so I've changed this to correspond with other snapshot files.

[renluop]

When I posted this morning, it was just before someone called for me. I misread the connection stats; apologies!

[renluop]

Now this is strange! Although I installed 5.8, now my graphs have 5.7.1 in the headings. I have never installed 5.7.1.
Moreover the text error averages have reverted to the old style, as can be seen from the attachment. Those earlier are in the new style.
WTH can be going on!

I must be the bain upon bains, bringing up all this! Sorry!

[roseway]

Unfortunately the .txt files in that zip are unreadable so I can't look at what you're describing. The only explanation I can think of is that you must have downloaded v5.7.1 at some time and you're running that.