[DRAFT] luci-app-statistics: DSL graphs update #6798
Conversation
|
So does attenuation ever become negative (i.e. amplification)? Does the relevant graph depend on its position in an array? Hmm. This is from a hh5a? Not to derail this discussion, but how was it to install that device? Did you require serial console as per the wiki? |
No, to my knowledge it could never be negative. This is why I'd suggest it's good to plot both up and down on the same graph, with up simply being flipped to the negative side.
Neither could I. I've basically just left uptime as it was before, with addition of some small formatting tweaks.
How you mean, sorry?
Yes, precisely. I've reflashed multiple HH5a's (or rebranded clones) by soldering directly to the test points on the board. |
There was a problem hiding this comment.
Can you explain the reason for these changes? To me these are just style changes, not fixes, and other files in the same folder use the formatting this file originally had: use " for strings, not have a whitespace betwen function and (, not use a trailing comma on the last item in a list.
Did you editor auto-format and introduce those changes?
|
Admittedly, yes this PR was thrown together rather hastily. There were a lot of stylistic changes in there which were enforced by my IDE's auto formatting. I have just reverted the single quotes and random whitespace. Hopefully that's a bit better and makes the diff more readable. |
|
Hmm, after a bit more thought, perhaps uptime can simply be displayed as a clock, years, months, days, (weeks?), hours, minutes, seconds, rather than a graph whose limit is axis height will effectively never change shape, only the axis units crammed in will get further between. I don't think rrdtool will do this, but just a large rendered string at top or bottom would be fine. |
|
Not sure I agree to be honest. It can be useful to see the history of when the line previously went down, which a simple counter wouldn't show. However, I have given it a thought and yes you're right, we need something better than graphing uptime in seconds. Newer versions of rrdtool have the ability to format a time duration into a number of days, hours, minutes, etc. See the docs. This would be great, but we're still stuck on v1.0.x (newest is v1.8.0, not sure when the time/duration formatting was introduced). I have opened an issue in the packages repo: openwrt/packages#23087. |
|
I think things like this were noted earlier issue #6139 - some of those are not simple to fix.
A counter would reset to zero (including everything else on platforms that don't save the graph data to storage), but it's not obvious without seeing the line start from zero again. |
How about instead of uptime, just graph line up status, which is basically boolean. That clearly shows when the line is not up. |
I'm still on 22.03, due to the issues with the HH5a, hence I cannot properly test this. My current test setup uses a collectd-exec and a shell script I hastily threw together which calls ubus to gather the DSL metrics. Crucially, not the same as the actual lua script now in the collectd package, so there is plenty of room for things to go wonky.
Not sure if the graph titles work correctly. %pi is meant to be plugin instance. For what I threw together using collectd-exec ages ago, this is
dsl0. I'm not honestly sure how the lua script does this.I think I have added all the necessary changes to the lua script for gathering the extra data, but if someone could test that would be greatly appreciated. openwrt/packages#23011
Paging @bluebrother because I know I'm rewriting much of their code.
Screenshots:
Not that there's much to show for the errors lol. Nor flags, my shell script didn't really deal with bools well