#linuxcnc-devel Logs

May 25 2017

09:02 AM skunkworks_: wow - my google foo is failing me.. I wonder if there is a way to see what drive a logical block belongs to in an MD array
10:31 AM cradek: skunkworks_: I'm so sad you/we don't have zfs in linux
11:15 AM skunkworks_: cradek, will it ever happen?
11:15 AM mozmck: run ubuntu! ;-)
11:17 AM cradek: freebsd is more to my liking that ubuntu. did they just disregard the license problem and dare someone to sue them?
11:19 AM mozmck: I haven't followed it really, but I've seen articles like this and it looks like a bit of a grey area: http://www.pcworld.com/article/3061924/linux/ubuntu-1604s-support-for-the-zfs-file-system-may-violate-the-general-public-license.html
11:41 AM jepler: cradek: I don't know if "dare" is the right word. They say they got legal advice that said it was not license infringement.
02:10 PM skunkworks: jepler, http://electronicsam.com/images/matsuura/20170525_132343.jpg
02:10 PM jepler: skunkworks: looks very nice
02:10 PM jepler: have you tried modeling anything yourself yet?
02:11 PM skunkworks: still amazing.. Definitely went into this with pretty low expectations..
02:11 PM skunkworks: jepler, no. not really
02:12 PM jepler: is that a printed gear on the right?
02:13 PM skunkworks: yes. for a nautilus thing.. (but the carrier broke)
02:14 PM skunkworks: The white seems a lot more brittle than the blue filament I have.
02:14 PM jepler: bummer
02:15 PM jepler: the hatchbox pla white is pretty similar to the hatchbox colors, but I have a different brand of "neutral/clear" pla which is more matte and maybe more brittle
02:16 PM jepler: a tactile difference too
03:50 PM jepler: the graphs of cpu and memory usage via the digitalocean control panel are sure more stable in the last ~20 hours (since I made that last round of changes)
03:50 PM jepler: memory usage reported there slowly crept from 33% to 50%, CPU usage is consistently under 10% it looks like
03:54 PM seb_kuzminsky: that sounds promising
03:54 PM seb_kuzminsky: do they give you io utilization? i think we're io bound, not cpu bound, on flo
03:54 PM jepler: no, they give disk I/O in MB/s which is dominated by some spikes at >200MB/s in the 7-day view
03:55 PM seb_kuzminsky: that seems hard to believe
03:55 PM seb_kuzminsky: or am i out of touch with modern disk speeds?
03:55 PM jepler: on the 6-hour view, typical read is <.1MB/s, typical write is <0.5MB/s
03:55 PM jepler: they claim they are all SSD
03:55 PM seb_kuzminsky: 99%+ going to those bunk cache files, i bet
03:55 PM jepler: or swap
03:57 PM jepler: looks like before the changes yesterday, typical write rate was >1MB/s
03:58 PM seb_kuzminsky: is the forum feeling slower after that change, does anyone know?
03:59 PM jepler: dunno
03:59 PM jepler: the docs I saw claimed that the cache was only applicable to not-logged-in sessions anyway
04:00 PM jepler: so the users we care most about, logged in ones, shouldn't see a performance degradation since they didn't get cached pages anyway
04:01 PM seb_kuzminsky: before the change yesterday we had extended periods (hour+) of swap thrashing and < 10 requests served per minute
04:02 PM seb_kuzminsky: after the change we've not had any of those, we've consistently served tens to hundreds of requests per minute, and almost no swapping
04:07 PM jepler: Server uptime: 22 hours 9 minutes 12 seconds
04:07 PM jepler: Server load: 0.10 0.14 0.13
04:07 PM jepler: Total accesses: 98099 - Total Traffic: 3.2 GB
04:07 PM jepler: CPU Usage: u5340.9 s372.71 cu.24 cs1.13 - 7.17% CPU load
04:07 PM jepler: 1.23 requests/sec - 41.8 kB/second - 34.0 kB/request
04:07 PM jepler: 6 requests currently being processed, 2 idle workers
04:07 PM jepler: apache has an internal status page
04:08 PM jepler: frequntly there are 0 idle workers, I wonder about bumping it up from 8 a bit
04:13 PM seb_kuzminsky: maybe a little
04:14 PM seb_kuzminsky: at some point it's better to leave the requests in the queue for the limited number of workers to deal with later, rather than having more workers running in parallel and sharing scarce system resources like memory and io bandwidth
04:14 PM seb_kuzminsky: but of course you know that, and i don't know where the break-even point it
04:14 PM seb_kuzminsky: *is
04:23 PM seb_kuzminsky: this is what flo's bonks looked like: http://forum.linuxcnc.org/tmp/requests.png
04:28 PM jepler: logging in via the administrator interface sure isn't going fast..
05:51 PM Connor1 is now known as Connor
10:06 PM Connor1 is now known as Connor