- This topic has 152 replies, 16 voices, and was last updated Jul 8-5:18 pm by Ampersand.
-
Posts
-
anticapitalista
I don’t think we have any runit experts. I’m not. If there is one, please help us out here. Thanks
Philosophers have interpreted the world in many ways; the point is to change it.
antiX with runit - leaner and meaner.
Xecure
I installed iotop to check read/write on HD.
sudo iotop displays what you can see in the image above.It is OK on rnit live USB (and it works fast with root persistence), but I consider it bad (compared to sysvinit) on an installed system. Writing ~500 KB/s may seem very little, but it takes it toll. and HDD will continually make noise, which I think will make it heat up a bit.
runsv (the service handler) is the one to blame. I will later check if I can disable logging and see if this activity stops.antiX Live system enthusiast.
General Live Boot Parameters for antiX.anticapitalista
It might also be down to our old ‘friends’ ntp and/or connman …
Philosophers have interpreted the world in many ways; the point is to change it.
antiX with runit - leaner and meaner.
Xecure
I cannot seem to find the way to stop all the runsv logging.
I see that iotop shows processes like
runsv acpi-support
that when I check the sv status, it says they are down. SO something is calling them and keeps writing who know what, probably a log, but I cannot figure it out at all.
When checking htop (trying to figure out what are the processes with “log” in their name), I can see a process that says:runsvdir -P /etc/service log: _runit-log:adm’ chown: invalid user: ‘runit-log:adm’ chpst: fatal: unknown user/group: _runit-log chpst: fatal: unknown user/group: _runit-log chpst: fatal: unknown user/group: _runit-log chpst: fatal: unknown user/group: _runit-log chmod: cannot access ‘/var/log/runit/acpid/*’: No such file or directory chpst: fatal: unknown user/group: _runit-log chpst: fatal: unknown user/group: runit-log
Which seems to be related to the runsvdir log, but I cannot find what files are being continually written. I am not sure if this is related or not to the current problem.
When I do
sudo lsof | grep "runsv" > runsv-opened.txt
I can see all files related to runsv that are currently open (if I am not mistaken on how lsof works). But this also doesn’t help e figure out how to disable the log or figure out why services that are down are still being written to.I will explore on a live-sub tomorrow to see if things are different in live compared to installed.
antiX Live system enthusiast.
General Live Boot Parameters for antiX.anticapitalista
Remove/delete the folder (via root rox-filer) or unlink /etc/service/* where * is name of service eg ntp.
Don’t remove the following though
dbus
elogind
all getty
slim
udevdPhilosophers have interpreted the world in many ways; the point is to change it.
antiX with runit - leaner and meaner.
Xecure
Thanks, anticapitalista. That reduced a lot the amount of HD writing (to 100 KB/s). Is there a better way to stop the continuous writing?
I would like to compare with other distros that use runit (I will try void linux tomorrow on a VM), just to figure out if this is normal runit behavior. If it is, then we should recommend runit for liveUSB (with dynamic persistence) instead of installed (at least I will follow this advice). I am too spoiled by the lightness and frugality of the perfect antiX linux you have designed. runit is a great speed improvement, but I am too lost and don’t know yet how to handle it.
antiX Live system enthusiast.
General Live Boot Parameters for antiX.Anonymous
Is there a better way to stop the continuous writing?
Your earlier comment “500 KB/s […] and HDD will continually make noise” left me wondering:
noise == just spinning (i.e. staying active, in defiance of a command ‘hdparm -y /dev/sda1’)
or
noise == movement of the heads due to actual writingIf the latter, how is this even possible?
Does this particular kernel configuration specify some (any?) non-implicit default mount.ext4 settings ?
(probably not)Is any mount handler in play which is overriding the implicit default “sync” mount.ext4 option (vs “async”)?
(probably not, er, I can’t recall finding such in the stock antiX configurations)
________________________Broader than the scope of this runit -centric topic:
In a live enviroment, I have observed that unless/until a “sync” command is issued,
writes to any of the ext4 devices are, indeed, deferred.
The delay interval is 15 -ish seconds, which matches the implicit value of the “sync” mount option.The persist-save script (and live-remaster) takes into consideration this likely (writing, flushing) delay.
Subsequent to the rsync operation, it explicitly issues a sync command prior to unmounting the device.Recently, I checked… and did not find ANY script which passes “async”
(nor any other explicit mount option) when calling the “mount_if_needed()” routine.To avoid polluting the ‘runit’ thread, I’ll followup in a separate topic to discuss the mount options
topic: async mount option, and data=journal (re: persist-save and live-remaseter)
https://www.antixforum.com/forums/topic/async-mount-option-and-datajournal-re-persist-save-and-live-remaseter/ModdIt
@skidoo,
constant disk activity on an installed a2 Runit Bullseye was what quickly alerted me to there being a problem.I was able to reduce the symptoms by giving conmann a timeserver list.
(known problem, patch is upstream, debian maybe, discussion and patch on conman site).Will try and find time to boot up the core 2 box with HDD Runit a2, install, then see if there is repeated head
actuation after conman calms down, that takes a while then it should query net time at about 20 minute intervall.
That box has a very audible/clicky fast 7200RPM diskRegarding runit, it looks, to my very limited knowledge like runit tasks are constantly trying to log but unable
to setup the files from control code due permissions.
The local log files in etc are only a few bytes, messages seem to point that way too.For now will only run live in daily usage, apart from the noted issues the a2 is very usable, running on my newest
computer, 12 years old core I5 quad it has proved excellent for learning/compiling and building some smaller new
packages I wanted to try.All the extra no longer needed dependency and source code just dissapears if shutdown without a save on a manual
save persistence stick. Fails are just a case of a no save reboot and start again from a clean basis..
andyprough
python3 is not sym linked to /usr/bin/python by default. I don’t know if that’s the way you intended, but I download youtube-dl and update it manually from github, and it complains if I don’t make the sym link. So, this isn’t really a bug report unless this is undesired behavior for the devs, in which case I guess it is a bug report.
Xecure
Your earlier comment “500 KB/s […] and HDD will continually make noise” left me wondering:
noise == just spinning (i.e. staying active, in defiance of a command ‘hdparm -y /dev/sda1’)
or
noise == movement of the heads due to actual writingIf the latter, how is this even possible?
Sorry for not making it clear. I mean that antiX Bullseye a2 runit edition installed (not live) is continuously writing to HDD. On a normal live USB + dynamic root persistence works perfect without any HDD writing. antiX-runit seems to work faster on both installed and live, but the strange writing behavior is not what I want on my installed system.
On live USB I usually only save changes or manually or at the end of the session (semi-automatic), so nothing will be writen on the USB if I don’t want it so.
antiX Live system enthusiast.
General Live Boot Parameters for antiX.I cannot seem to find the way to stop all the runsv logging.
^—v
Remove/delete the folder (via root rox-filer) or unlink /etc/service/* where * is name of service eg ntp.
Don’t remove the following though
dbus
elogind
all getty
slim
udevd>>> Remove/delete the folder
Not on-the-fly though, right?
…and, certainly not while an affected service is underway and is actively writing.Doesn’t the necessity of resorting to altering the O/S directory hierarchy, on-the-fly, represent an inherent (and asinine) design flaw?
In an earlier topic, I posted a link (and screenshots) to for an “sv manager gui” program.
Today, I revisited to check whether it handles toggling of logging for individual services.
No, it does not, but
https://github.com/daltomi/xsv/blob/master/src/xsv.cpp#L1073
the program could be modified to additionally handle that task. (stop service, if running; mv directory; restart the service if it was running)anticapitalista
runit-on-antix-bullseye-iso might help
Philosophers have interpreted the world in many ways; the point is to change it.
antiX with runit - leaner and meaner.
Service directory structure
This is a tree of a complete service directory structure (aka /etc/sv/servicedir),
in some run scripts, typically only run will be available as usually it’s the only file needed.servicedir
├── run (755)
├── check (755)
├── conf (644)
├── finish (755)
└── log (directory)
├── config (644)
└── run (755)Yeah, reviewing that helped to reinforce my impression that izta “poor design choice, or outright design flaw”.
“Simplicity”, at the expense of pushing the burden of performing scandir and lsof and whatnot…
…onto the sysadmin’s (or helper app author) shoulders.Given that we are not dabbling in “embedded linux” or other one-time (permanent, unchanging) system setup… I’m at a loss to understand how can this be construed as “superior to” the construct in which a sysadmin edits declarative statements ONE canonical initscript per service, and the (self-described) “service manager” attends to the details of managing and effecting the changes.
https://en.wikipedia.org/wiki/The_Emperor’s_New_Clothes
“The townsfolk uncomfortably go along with the pretense, not wanting to appear inept or stupid, untila childskidoo blurts out that the emperor is wearing nothing at all. The people then realize that everyone has been fooled.“Xecure
Maybe the error reported in htop may be related (something related to runit trying to access a file/folder but doesn’t have the correct permissions)
runsvdir -P /etc/service log: _runit-log:adm’ chown: invalid user: ‘runit-log:adm’ chpst: fatal: unknown user/group: _runit-log chpst: fatal: unknown user/group: _runit-log chpst: fatal: unknown user/group: _runit-log chpst: fatal: unknown user/group: _runit-log chmod: cannot access ‘/var/log/runit/acpid/*’: No such file or directory chpst: fatal: unknown user/group: _runit-log chpst: fatal: unknown user/group: runit-log
We can see runit-log trying to do something, but it doesn’t have permissions (or cannot figure out how to write there).
I was hoping this could help figure out the problem, but I am no expert. I still think figuring this out is resolving an extra bug, just to make runit even more stable.
- This reply was modified 2 years ago by Xecure.
antiX Live system enthusiast.
General Live Boot Parameters for antiX.
Brian Masinick
I’m mot sure if this provides any direct solutions, but there are discussions regarding runit log files, the nature of their creation and use, a note that they are not mandatory (though recommended) and there are a couple of links to other resources that may assist in learning and managing log files. Like others, I am definitely NOT an expert on this either, but I happened to spot this discussion:
https://serverfault.com/questions/626436/how-do-you-configure-runit-logging
--
Brian Masinick
- You must be logged in to reply to this topic.
