118[01:58:59] <nvz> not when it comes to finding something
119[01:59:07] <nvz> you search for system tray you're gonna
get windows apps
120[01:59:30] <anonzzz> touché
121[01:59:49] <nvz> there are frameworks that seem to have
similar promising utility as the notification area but very little
uses them so far and they didnt seem very well documented
122[01:59:54] <nvz> I speak of things like appindicator
123[02:00:13] <nvz> in the future we may see more ease in
putting anything into such a thing
124[02:00:35] <nvz> in fact I think some panels like xfce's
have applets that can run a command and put its output into the
terminal
125[02:00:56] <nvz> with such a thing it'd be trivial to
display the information in question in a panel
126[02:01:02] <nvz> just not in the notification area
162[03:05:02] <JordiGH> Where does system store journals?
163[03:06:37] <sponix> JordiGH: for systemd ? or for damn near
everything else ?
164[03:07:16] <JordiGH> y'know what, I'll just strace
journalctl
165[03:07:31] <Unit193> Either in memory or /var/log/journal/
166[03:08:09] <Unit193> sponix: Heh, and just now a new esr is
uploaded.
167[03:08:35] <sponix> Unit193: you are freaking kidding me --
right ?
168[03:09:03] <Unit193> To unstable, but still.
169[03:09:17] <sponix> Unit193: its been in "Sid" for
a bit
170[03:09:27] *** Quits: amlchief (~amlchief@replaced-ip) (Remote host closed the connection)
171[03:09:38] <sponix> I'm doing up a VM to see what is
required to backport it to Buster
172[03:09:46] <sponix> Might be way more deps than I want to
deal with
173[03:09:50] <sney> oh hey, 78.3.0esr-1 with an upload date of
September 23rd. It's still the 22nd where I am
174[03:10:07] <sney> now that's timely.
175[03:10:42] <JordiGH> So what's /run/log/journal?
176[03:10:46] <JordiGH> I have no /var/log/journal
177[03:10:54] <JordiGH> I just found /run/log/journal by
stracing.
178[03:11:34] <sponix2ipfw> is it judd or dpkg that does the ssb
?
179[03:12:06] <sney> sponix: building firefox is a large
undertaking. if you succeed in building it for your own purposes,
then you'll have it for a while before the rest of us. but
ff's maintainer will certainly get it into buster-backports
before long
180[03:12:16] <sney> dpkg does ssb, judd does checkbackport
197[03:16:06] <JordiGH> sney: Does journalctl work without root
in daredevil?
198[03:16:12] <SerajewelKS> the fuse exfat driver can barely
maintain 2MB/s write throughput and pegs a CPU core while it does
199[03:17:09] <sney> JordiGH: I don't know what you mean by
daredevil, but you still need the systemd-journal group for normal
user to access systemd's journal in bullseye, as in buster.
201[03:17:26] <JordiGH> Ha, sorry, I meant bullseye.
202[03:17:34] <JordiGH> My brain did an accidental substitution.
203[03:17:56] <sney> the main difference is that /run is a tmpfs
while /var is not. (by default, anyway.) there may be a
justification for that change in bullseye's systemd's
changelog.
208[03:21:16] <sponix2ipfw> Unit193 sney : after seeing the
level of involvement to build firefox-esr --- I'm just gonna
wait for it to hit the backports. My slightly dated version is
working just fine LOL
209[03:22:15] <SerajewelKS> looks like
replaced-url
210[03:22:24] <sney> SerajewelKS: not available through the
usual debian sources. some modules can be backported from one kernel
to another, though it's not a trivial thing. if exfat.ko exists
in a late 4.x kernel that might be doable
211[03:22:43] <sney> it'll be a fun hacky project, if you
enjoy that kind of thing
214[03:23:13] <SerajewelKS> sney: it looks like the source i
just pasted should work, but the dkms package there is for ubuntu,
so i can either try that and cross my fingers or i can build it with
make.
215[03:23:57] <sney> a dkms package from ubuntu would probably
just work, or at least with minimal modification
216[03:24:39] <SerajewelKS> that was my thought as well. but
adding an ubuntu PPA to debian makes me itch.
217[03:24:57] <sney> ubuntu PPAs have .dsc files like any debian
repo, so you can grab that manually with dget
237[03:45:38] <sney> it's a kernel module from a newer
kernel backported by one rando on github, it's basically
pre-release software and should be considered unreliable
238[03:45:57] <sney> awesome that it built, though :P
270[04:01:57] <SerajewelKS> indeed. but my backup drives are
connected over USB and i don't want my backups failing. with
sustained writes on a crypt layer on top of a USB drive, i would
suddenly get a USB disconnect and the drive would go offline until
replugged.
271[04:02:10] <SerajewelKS> sounds like a drive problem EXCEPT
it was consistent with 5.5 on every device i had available to test
274[04:03:14] <SerajewelKS> i never found any mention of the
issue so i have no idea if it's fixed in 5.7. i also use btrfs
and it's strongly discouraged to use a btrfs volume on a newer
kernel then go back to an older kernel.
275[04:04:58] <sney> well that's a zfs feature I would have
copied, sheesh
279[04:06:59] <sney> newer zfs versions support more features,
but you have to deliberately upgrade your storage pool for them.
until you do that, the pool is still usable by the older versions
280[04:07:09] <SerajewelKS> maybe that's specific to raid5
(which is still very much in development)
281[04:07:44] <SerajewelKS> btrfs has similar compat flags
386[05:53:46] <edufmass> Hello, I unziped an app into
/opt/appname and I run /opt/appname/bin/app.sh, How to enable to
just run app command everywhere?
391[05:59:43] <sney> /usr/local is not touched by apt or dpkg.
so anything you put there is not going to conflict with a debian
package, as long as it doesn't have the same name (eg
'python') as an important binary.
417[06:13:00] <sney> sponix: I've been on an ipsec vpn
configured on my router since february, and it's extremely
reliable, so I'll probably wait until some news from either
vendor tells me it's time to switch. regardless, thank you for
the link, but I am a retired networking professional.
419[06:14:44] <sponix> sney: I only know what networking is
needed for server/client/rig administration .... most of which was
learned just by "tinkering"
420[06:15:19] <sney> that is a good place to start.
421[06:16:02] <sney> and a lot of places will only hire you if
you know how to research a solution to an issue that you
haven't heard about recently.
422[06:16:18] <sney> um, but once you get paid for it, you are
generally pretty handy with VPNs.
423[06:17:14] <sponix> sney: I couldn't figure out what
config went in what box on the Mobile Android Wireguard app LOL ...
So, I hand jammed the config on my rig, and used a QR code generator
to scan the config for import :)
424[06:17:37] <sponix> sney: _now_ I can look at those config
boxes though, and see what I should have been entering where
425[06:18:34] <sponix> have to admit. It took me about an hour
and a half to learn how to do the basic configuration of both the
server and client -- But most of that time was spent searching to
find this guide that was worth a shit
426[06:18:45] <sponix> research is half the battle though
428[06:19:54] <sney> I try to filter the guides and forums out
of my search results as much as possible.
429[06:21:01] <sney> everyone ships a manual of some sort. the
more complicated stuff includes examples. the rest is
"tinkering" as you said.
430[06:21:40] <sney> whoever wrote that guide or shot that
youtube video had to figure it out too.
431[06:23:16] <sponix> other guides/examples were either more
complicated than they needed to be, or only covered some aspects of
what I needed. This one is segmented, but at least all segments
coincide
432[06:23:55] <sponix> linuxbabe.com has some pretty good guides
:P
570[09:40:06] <ariejan> Hi. I'm running buster 10.5 in lxc
on proxmox. The tmpfs mount for /run/lock is set to 5M which is
insufficient for my current samba install. I want to resize this,
but my added fstab entry seems to get ignored completely. Any ideas
how to change the size of /run/lock?
573[09:41:51] <dpkg> Proxmox Virtual Environment (Proxmox VE) is
a GNU/Linux distribution <based on Debian>, providing a
virtualization platform with <LXC> and <KVM>. It is not
supported in #debian. There's an unofficial
proxmox channel on Freenode. For official venues, see
##replaced-url
574[09:42:25] <ksk> ariejan: I can imagine you would need to
change that on the host, might be its not "virtulized".
575[09:42:41] <ariejan> @ksk hmm, good point. I'll look
into that.
576[09:42:42] <nvz> ariejan: the tmpfs are likely setup in early
boot stage probably in initrd even
578[09:43:04] <ksk> (inside lxc, there is no (additional)
kernel)
579[09:43:38] *** Quits: Tobbi (~Tobbi@replaced-ip) (Quit: My MacBook has gone to sleep. ZZZzzz…)
580[09:44:51] <nvz> I would imagine you can probably in debian
set something like /etc/default/tmpfs and rebuild the initrd but
this is all irrelevant in this scenario
603[10:02:46] <jelly> ariejan, you can mount -o remount,size=10M
/run/lock on the fly to test whether it's going to work at all;
I have no idea how limited LXC containers are and whether
you're allowed to mount and umount things
609[10:16:33] <ksk> You cant, but chances are you can do so from
the host (or modify the container itself, so it gets done on startup
(via lxc.config) )
649[11:25:03] <j2mb0> hey guys, yesterday i was struggling to
install audacious of debian testing but with ur help i finally
managed to do it after purging the underlaying broken libraries,
reinstalling them then installing audacious. thank you again
650[11:26:02] <j2mb0> my next problem is trying to install
dockbarx, my terminal outputs "depends: python: any(<2.8)
but it is not installable. I wonder how to start researching the
subject, what does this output mean?
651[11:28:12] <ksk> !debian-next
652[11:28:13] <dpkg> #debian-next is the channel for
testing/unstable support on the OFTC network (irc.oftc.net), *not*
on freenode. If you get "Cannot join #debian-next (Channel is
invite only)." it means you did not read it's on
irc.oftc.net. See also
replaced-url
653[11:28:48] <ksk> j2mb0: that means that this package needs
python in a version prior to "2.8" to run - and that there
is no such python available for your debian release
654[11:28:54] <j2mb0> i dont think the subject is exclusive to
testing
655[11:29:01] <ksk> !frankendebian
656[11:29:01] <dpkg> When you get random packages from random
repositories, mix multiple releases of Debian, or mix Debian and
derived distributions, you have a mess. There's no way anyone
can support this "distribution of Frankenstein" and
#debian certainly doesn't want to even try. Ask me about
<reinstall>
657[11:29:03] *** joze is now known as jumijoze
658[11:29:29] <j2mb0> ksk: and why isnt it installable if it
needs older versions?
659[11:29:36] <ksk> j2mb0: of course it is, stable has different
versions of anything, and you -should not- ever encounter such an
error in stable.
660[11:29:58] <ksk> because its dependencies are not
fullfillable by apt
661[11:30:06] <j2mb0> How can i allow an exception?
662[11:30:10] <ksk> You cant.
663[11:30:34] <ksk> You should not "just run testing"
btw. Chances are high you already broke your system in various ways.
664[11:30:56] <j2mb0> So it could be my system and not the apt
sources?
665[11:31:27] <ksk> !bat
666[11:31:27] <dpkg> [Basic Apt* Troubleshooting]. To diagnose
your problem, we need ALL OF THE FOLLOWING information: 1. complete
output of your apt-get/apt/aptitude run (including the command used)
2. output from "apt-cache policy pkg1 pkg2..." for ALL
packages mentioned ANYWHERE in the problem, and 3. "apt-cache
policy". Use
replaced-url
667[11:31:53] <j2mb0> what output are you looking for? "apt
install dockbarx"?
668[11:31:56] <ksk> eh, nevermind. If you are running testing,
you need to go to #debian-next on OFTC network. No need to provide
the "BAT" information in here..
669[11:32:46] <ksk> Feel free to link the output over there. And
no, we would need everything the faqtoid just told you, not just one
out of X.
670[11:33:12] <j2mb0> i wont be able to enter the channel if it
is invite only
671[11:33:13] <ksk> But I can only repeat myself: If you are not
familiar with (Debian) Linux, you should not run testing.
672[11:33:30] <ksk> j2mb0: read the instructions about debian
next again. Hint: Its on another network.
674[11:34:06] <Lope> is it possible to add a wifi adapter,
connected as a "client" to a bridge, with network-manager?
675[11:34:07] <j2mb0> I thought i am familiar with linux
dependencies until i realized this problem is infeasible
676[11:34:24] <j2mb0> thank you for your help, i will go to OFTC
677[11:35:32] <ksk> Did you have a reason to go to testing? If
not, reinstall :P
678[11:35:56] <j2mb0> I cant afford reinstalling right now, my
job is more important, but i will consider that
679[11:36:15] <ksk> Oof², running testing in production
is..
680[11:36:18] <ksk> !testing security
681[11:36:18] <dpkg> Security updates in testing are delayed by
the normal testing migration *and* may be further delayed by missing
dependencies, etc. See
replaced-url
682[11:36:22] <j2mb0> An expert friend suggested taking the
middle way not stable and experimental but testing :)
683[11:36:40] <j2mb0> It is my home computer
684[11:36:49] <j2mb0> good advice tho ;)
685[11:36:53] <j2mb0> thought*
686[11:37:04] <j2mb0> though* omg am confused
687[11:37:11] <Haohmaru> testing is not for n00bz btw
689[11:38:33] <Haohmaru> i was recommended to get debian
testing, but then i quickly found out that it's like a forest
at night... so i went for stable
690[11:39:10] <Haohmaru> in #debian-next they expect you know
wtf you're doing (and i didn't even..)
746[12:33:22] <Lope> flatpak remote-ls: "error: GPG
verification enabled, but no summary found (check that the
configured URL in remote config is correct)"
766[12:54:55] <Lope> does anybody know how to make
flatpak/flathub work?
767[12:55:06] <Lope> flatpak remote-ls: "error: GPG
verification enabled, but no summary found (check that the
configured URL in remote config is correct)"
768[12:55:15] <ksk> !problem
769[12:55:15] <dpkg> [problem] something that can be solved,
fixed or worked around if properly described. A good thorough
description of the problem, with detailed steps of how to reproduce
the problem, the produced output, and the expected output, is the
best start to discuss a problem.
866[13:58:29] <ix_> hello, can anyone tell me how to permanently
disable a service? I want to disable NTP, because there is an issue
with the time (dual boot)
872[14:00:00] <ratrace> ix_: systemctl disable ntp.service,
assuming ntp.service is the service unit for it :: there are several
ntpd programs so depends which one you have.
873[14:00:09] <hmuller> do these images:
replaced-url
875[14:00:35] <ratrace> also, you might wish to systemctl mask
<unit> because disable'ing only affects explicit
starting; other services may force a dependency start, so
"mask" is what you want to completely, fully, prevent it
from starting
889[14:05:53] *** Quits: _aeris_ (~aeris@replaced-ip) (Remote host closed the connection)
890[14:05:56] <ratrace> then you're probably running
systemd-timesyncd. verify with systemctl status
systemd-timesyncd.service
891[14:06:57] <ix_> yep, that's the one
892[14:07:05] <ratrace> well, you can't uninstall that. you
can "disable" or better "mask" the
systemd-timesyncd.service . assuming you actually want to shoot
yourself in the foot and disable an NTP service. clocks drift all
the time.
893[14:07:21] <ratrace> ix_: but eh.... dual boot with windows
issues? you can regedit windows to use UTC time
908[14:12:00] <ix_> yes, I know that the clock drifts, I'm
fine with it
909[14:12:18] <ratrace> ix_: in which case you might wish to set
your rtc time in debian to localtime
910[14:12:30] <ratrace> timedatectl set-local-rtc true (I think)
911[14:12:52] *** Quits: banisterfiend (~textual@replaced-ip) (Quit: My MacBook has gone to sleep. ZZZzzz…)
912[14:12:52] <ratrace> that way, windows time-sync will set rtc
clock to localtime, so at least you'll have somwehat correct
clock when you boot back from windows
913[14:13:11] <ix_> ratrace, my Debian clock is UTC
914[14:13:30] <ratrace> you'll have to set it to locatime
unless you want it off
915[14:13:43] <ratrace> off = offset by your timezone
916[14:14:17] <ix_> ratrace, well, if it thinks that my time is
UTC and the BIOS time is UTC, then it will show me BIOS time, which
is actually local time
917[14:15:03] <ratrace> my OCD is screaming, pulling its hair
out and screaming, now, btw, so thanks for that.
918[14:15:10] <ix_> :)
919[14:15:11] <hmuller> lol
920[14:16:00] <dvs> ix_: you want to disable ntp because the
BIOS clock keeps getting changed by Debian?
947[14:24:15] <no_gravity> Is there a command like
"watch" that does not clear the screen after each output?
So you can do "watch uptime" and see kind of a log of the
load?
948[14:24:53] <ix_> ok, so I did the following: timedatectl
set-local-rtc 1 && timedatectl set-timezone mytimezone
949[14:25:12] <ix_> it worked beautifully
950[14:25:20] <RadoS> no_gravity, you can do that yourself in
shell-cmds.
976[14:33:33] <Lope> ratrace, selecting Haswell / Haswell-IBRS /
Haswell-NoTSX / Haswell-NoTSX-IBRS in KVM is not about emulation.
it's about restricting instruction sets. And by doing so one
might get better perf because of not needing to do mitigations for
those instruction sets.
979[14:35:24] <ratrace> Lope: if you want to disable
mitigations, then disable mitigations in the guest (and host).
adding vt-x traps to instructions is not the way to go
980[14:35:50] *** Quits: Haudegen (~quassel@replaced-ip) (Quit: Bin weg.)
981[14:37:01] <Lope> ratrace, what do you mean by
"traps" surely selecting the instructions just limits what
the "cpu" will tell the kernel it has?
984[14:37:53] <ratrace> Lope: vt-x works by setting
"traps" or interrupts where the CPU will hand off
execution to a hypervisor-specified function
985[14:38:11] <ratrace> Lope: without that, the hypervisor would
need to run full emulation layer which would nuke the performance
986[14:38:21] <Lope> ok
987[14:39:08] <ratrace> also I'm not sure that by disabling
specific instructions you actually prevent spectre/meltdown
mitigations from running. maybe a small subset of them, but really
if you want them off, just disable them.
988[14:39:29] <ratrace> for linux guests, add
"mitigations=off" to kernel command line and bob's
your uncle. for windows, google for it, you'll have to regedit
some things, but it's doable.
989[14:39:39] <Lope> ratrace, I wonder why there are 4 variants
of Haswell offered by KVM?
990[14:40:26] <hmuller> is that "uncle=bob" ?
991[14:40:29] <Lope> mitigations=off sounds cool, but then you
better not browse the web in the VM etc.
993[14:40:55] <ratrace> Lope: the browsers fixed javascript
timing precision so it can't be used to xploit spectr or
meltdown
994[14:41:49] <genr8_> Those variants are hardware instruction
capability Mitigations=ON though
995[14:42:06] <ratrace> and also, I have yet to read about a
report of meltdown/spectre being actually used in attack in the
wild. they're very hard to pull off, and make sense only for
some limited, highly targeted server virtualization circumstances
996[14:42:34] <ratrace> "wrench security" would be
much, much, much cheaper with much better results, for any attacker.
997[14:42:51] <genr8_> I anticipate no real difference in
setting the flags in the hypervisor vs the guest kernel
1000[14:43:28] <genr8_> cloud providers would want to enable them
that way.
1001[14:43:48] <ratrace> genr8_: well some things are done twice,
like trampoline and other mitigations in gcc
1002[14:43:54] <genr8_> where they cant control the guest, and
they opt for mitigation instead of speed
1003[14:43:57] <Lope> ratrace, also, a spectre/meltdown attack is
probably quite obvious in terms of high CPU usage.
1004[14:44:20] <genr8_> the retpoline is different than the IBRS
hardware instruction provided through microcode
1005[14:44:28] <ratrace> Lope: maybe, maybe not. Have you ever
seen one? ;)
1006[14:44:40] <Lope> kind of like when you boot a windows 10 VM
and it Windows de(off)ender grinds 8 threads at 100% CPU usage for
15 minutes on every boot.
1009[14:45:31] <ratrace> Lope: it's much likely to contract
something much more efficient, than meltdown/spectre, on windows.
1010[14:45:38] <ratrace> much *more likely
1011[14:45:55] <genr8_> IBPB+retpoline or IBPB+IBRS Is needed to
fully mitigate it
1012[14:46:05] <Lope> ratrace, can't say I have. But it has
to hammer the fuck out of CPU registers and whatnot to extract one
bit per every few minutes or whatever, so I'm sure that would
involve some high CPU usage.
1017[14:47:42] <genr8_> like I said, those 4 Haswell choices are
really for cloud providers who control the hypervisor only. If you
control your own OS, you can do other things.
1027[14:49:47] <ratrace> Lope: second most secure option is
native CPU (no hypervisor shenanigans) with up to date microcode;
the first most secure being using a CPU that's not vulnerable
in the firrst place
1028[14:49:59] <jelly> Lope, I don't notice 15 minute loads
on our virt. enviroment _at all_; it has peg CPU and run for
hours-days for us to react
1054[14:57:18] <genr8_> yea, i just confirmed, you need to
explicitly set the "IBRS" cpu option in qemu to expose
that feature to guests. just having a safe CPU does not allow the
guest to take advantage of it
1055[14:58:11] <genr8_> heres a giant guide
replaced-url
1056[14:58:32] <ratrace> {g}: I do that for nginx, but quick
google shows me Apache can do that with access logs too. I log
response processing time, both total and specifically backend. That,
in correlation with client-side (eg in firefox, developer mode, you
can see various stages of request), you can get a much better
picture of where the problem is
1063[15:00:22] <ratrace> {g}: so you get logs showing how much
time it took to service the request in total (from connect to
close), how much of that was backend, how much of that was web
server (up until first byte is sent), etc...
1064[15:00:30] *** Quits: dvs (~hibbard@replaced-ip) (Remote host closed the connection)
1065[15:00:40] <ratrace> {g}: but of course it is helping. you
see exactly how much time particular stages take
1066[15:00:48] <genr8_> if you are picking the names, you have to
explicitly pick the IBRS name. if you are picking
"passthrough" then you don't.
1067[15:01:15] <ratrace> for example, if apache does everything
with 10msec, but total response time (which includes sending back to
the client) is 1000msec, that means 990msec was in the network,
sending back
1110[15:14:34] <genr8_> its not fine. you dont want the content.
1111[15:14:44] <{g}> genr8_: Why not?
1112[15:14:51] <ratrace> still.. that's higher on the OSI
stack so non-network processing is involved in that total time. not
quite comparable to lower level ICMP
1113[15:15:26] <ratrace> {g}: and you can't measure network
congestion with a single packet. you need flows, and correlate
latency with bandwidth with packetloss
1114[15:15:58] <{g}> ratrace: I can run the "time wget
..." multiple times to get a feel.
1124[15:19:34] <ratrace> also that curl test is bs. like genr8_
implied, getting back content includes server-side processing time
(though HEAD may do that too) and there's no way to isolate it
1125[15:19:35] <{g}> ratrace: Google is probably answering from a
CDN endpoint that is close enough to me.
1126[15:19:55] <ratrace> {g}: no. you're doing it wrong.
1127[15:20:00] <genr8_> its better than wget.
1128[15:20:14] <{g}> ratrace: I want to know if my datacenter is
shaping my traffic. For that, I think its fine.
1129[15:20:15] <ratrace> comparing apples to oranges, not
controling variables, your metrics are worth zero.
1130[15:20:41] <ratrace> {g}: most certainly isn't. your DC
could do _different_ QoS for different routes, eg. to google
1131[15:20:44] <genr8_> one get request is not gonna tell that
1132[15:20:58] <genr8_> you said yourself theres CDNs. and
different routes.
1133[15:21:18] <{g}> There is always a "could".
1134[15:21:35] <genr8_> internet exchanges and peering matters
also
1135[15:21:52] <ratrace> {g}: but you have no way of knowing
that, so your metric in _that_ respect is useles
1136[15:22:17] <genr8_> you want something like PathPing, or MTR
1137[15:22:22] <{g}> ratrace: Well, then we have different
opinions. I think we exchanged our argument.s
1138[15:22:27] <ratrace> eg. wget'ing 100b from google and
wget'ing 100b from your own server, the difference in time
tells you exactly nothing.
1139[15:23:02] <genr8_> ratrace is right
1140[15:23:03] <ratrace> {g}: except this is not a matter of
opinion but scientific methodology of isolating variables when
collecting objective metrics.
1141[15:23:31] <ratrace> if you get 20msec in one and 40msec in
the other you have NO other datapoint to tell why there's
20msec difference.
1143[15:24:24] <{g}> ratrace: If I see that the time to wget
google goes up 10x every time I see my pages take 10x longer, I can
assume it is because of the network.
1144[15:24:43] <ratrace> {g}: no. it could be due to a situation
in your kernel
1145[15:24:49] <ratrace> see? no datapoints to tell the
difference.
1147[15:24:59] <{g}> If google stays at ~10ms even when my own
pages suddenly take 10x longer, I can assume it is not the network.
1148[15:25:18] <ratrace> quite the contrary, it could be the
specific route to your server. no datapoint to tell otherwise.
1149[15:25:50] <genr8_> thats why you want the ICMP ping, not a
full HTTP request
1150[15:25:58] <ratrace> {g}: unless you sit on a backbone
router, there's probably 5-10 hops between your machine and
your server. more if your server is geographically distant
1193[15:43:42] <ratrace> genr8_: what's that ubuntu page
actually saying, I'm not sure I get the context of why you
posted it. does it mean native cpu in qemu-system won't
actually expose all instructions to the guest?
1195[15:46:01] <genr8_> yes. if you run qemu with the named CPU
models, it won't.
1196[15:46:13] <genr8_> you have to run it with the "Host
passthrough" model, aka "-cpu host"
1197[15:47:09] <genr8_> if you use the names, the non-IBRS is
"legacy" and was not updated. a new Name was added for
IBRS instruction support. (so as not to break peoples stuff)
1247[16:06:12] <ratrace> dupin: so what's the actual
problem?
1248[16:06:23] <greycat> zerocode: show us the ACTUAL ERROR that
inspired you to try to purge this package. Then show us the ACTUAL
COMMAND that you typed when attempting to purge it, and the ACTUAL
ERROR from that command.
1249[16:06:26] <dupin> I cann not open thunar
1250[16:06:37] <ratrace> dupin: try it from command line, see if
it complains about anything
1262[16:08:05] <genr8_> good. he has broken brain. purged too
many brain cells
1263[16:08:08] <ratrace> I really shouldn't laugh. some
people are genuinely dislexyc. not a laughing matter. but I can
laugh at their choice to use text-only medium for help? yea, I can,
can't I.
1329[16:36:38] <genr8_> i found out the hard way that hostnames
in iptables scripts cant be resolved on startup with no internet
connection and cause resolve errors and cancel out the entire script
from loading
1330[16:36:49] <jolt> I'm out of batteries on my remote
mindreading cap, so I can't help either.
1379[17:03:37] <match_it> ivzhh, I "remember" yes but
none of broken
1380[17:03:38] <genr8_> that would be why :)
1381[17:03:52] <genr8_> buster removed a lot of packages, that
seems to be one of them
1382[17:04:03] <greycat> I'm also really suspicious of how
the reported errors are truncated right where any useful information
like the package version might appear.
1389[17:05:59] <greycat> The package version number probably says
something like "ubuntu" in it, which would be why a
malevolent person would try to hide it while asking for help in the
wrong channel.
1390[17:06:18] <ksk> Sora: Ignore that, gimme a second, wrong
link..
1393[17:06:47] <jelly> Sora, sadly yes, the linux kernel is still
a bit buggy when there's no swap at all
1394[17:06:49] *** Quits: kale (~kale@replaced-ip) (Ping timeout: 258 seconds)
1395[17:07:00] <jelly> give it just a tiny bit like 64MB or 512MB
1396[17:07:10] <ratrace> Sora: abd tgus
replaced-url
1397[17:07:13] <jelly> it helps in corner cases
1398[17:07:15] <cdown> Sora: Swap space and RAM are mostly
tangential. Swap itself is more or less essential in any modern
operating system, because you need the ability to promote forward
reclaim for types of memory which are otherwise locked in
1399[17:07:16] <ratrace> lol! *and this
1400[17:07:44] <cdown> I'd recommend 512M+, more if you
happen to have a lot of anonymous slack.
1425[17:10:21] <jelly> cdown, I solve that by regular swapoff -a;
swapon -a :-(
1426[17:10:38] <ivzhh> if you make the swap on lvm, then you can
adjust later.
1427[17:10:41] <cdown> Sora: The kernel doesn't swap without
reason, so if you have space, give it a few gigabytes. If
you're space-constrained, then whatever isn't going to
result in it filling up and thrashing.
1428[17:10:54] <ratrace> cdown: if swap gets full you need to
rethink your configuration, RAM size or (over)commitment policy
1429[17:10:58] <Sora> nah I could give it 128GB or whatever
1430[17:11:01] <Sora> but I don't think there's a
reason
1431[17:11:08] <cdown> ratrace: Or you need to rethink your swap
amount.
1432[17:11:11] <cdown> jelly: lol, did you know why swapoff -a is
so slow? I've still got that on my list to fix
1433[17:11:12] <jelly> because even with vm.swappiness = 1 the
thing still manages to fill it up sometimes even if there should
never be that much pressure
1437[17:11:31] <ratrace> cdown: if your system needs to swap in
and out at, say, 32G, even with SSDs that's gonna slow down the
system significantly
1438[17:11:46] <ratrace> the more swap you give it, the more pain
it's gonna inflict in swapstorms
1439[17:11:54] <jelly> cdown, it is absolutely not optimized at
all. I suspect the order it goes is basically random, and I/O are
page-sized-tiny
1440[17:12:14] <jelly> cdown, it has worked this way since like
kernel 2.4
1441[17:12:17] <ratrace> so, you need swap just to take off the
edge and satisfy buggy kernel mechanics that _rely_ on some of it
existing. anything else needs changing your RAM size or how much you
take from it
1442[17:12:25] <cdown> ratrace: If _that_ happens, it means you
don't have enough memory. But there are plenty of other reasons
you might run out of swap space.
1443[17:12:29] <cdown> For example, having tons of slack.
1444[17:12:35] <cdown> jelly: Oh, it's way worse than that.
1445[17:12:51] <cdown> jelly: We do an O(n^crazy) walk through
each virtual memory area, for each page, with no caching whatsoever
1446[17:12:58] <cdown> That's why swapoff is often CPU-bound
1447[17:13:09] <ratrace> cdown: the only reason you're
running out of swap space is that you have less memory than your
applications want.
1448[17:13:12] <Sora> so large amounts of swap is actually bad?
1449[17:13:21] <ratrace> there is no other reason. any other
reason can be distilled back to that one reason
1450[17:13:29] <jelly> cdown, so it does things in... memory
address order? :-)
1451[17:13:39] <Sora> i know this isnt debian but on windows i
remember i didn't have enough swap and the system used to run
out of memory even though it was only using 50% of the actual ram.
dont know if that applies to linux
1452[17:13:40] <cdown> ratrace: That's simply incorrect,
I'm afraid. Applications and userspace memory allocators often
simply do not know how much they need ahead of time.
1467[17:16:23] <ratrace> cdown: or in other words you can
overcommit way more than physical RAM (hence the name), because at
the end of the day it's all about how many pages are active and
what the kernel thinks can page out as inactive
1468[17:16:32] <cdown> ratrace: Ok, but there can be absolutely
humongous amounts which are only needed at some tiny fraction of the
process lifecycle.
1469[17:16:34] <jelly> except if kernel bits do crazy stuff
(nvidia.ko + zram+swap = bad things)
1470[17:16:49] <ratrace> cdown: well, yes
1471[17:17:03] <cdown> ratrace: So your statement that the reason
cannot be "you need more swap" is simply incorrect
1472[17:17:15] <ratrace> jelly: or you throw in ZFS into the pot.
THEN stuff gets extra crispy, juicy, interesting and mindblowing, at
the same time! :)
1473[17:17:16] <dob1> I am reading sed manual, I don't
understand /pattern/d pattern is a regexp?
1474[17:17:17] <cdown> In many cases, slack for workloads can be
in the order of tens of gigabytes.
1476[17:17:35] <ratrace> cdown: you mean "need more
RAM"? that's what I said
1477[17:17:38] <cdown> And that's production workloads which
are already optimised for physical memory allocation. For non-prod,
it's definitely a lot worse.
1478[17:17:41] <genr8_> sed /d is delete
1479[17:17:45] <jelly> ratrace, well ok, ZFS has Solaris VM bits
bolted onto Linux, or something
1481[17:17:53] <cdown> ratrace: 16:13:09 < ratrace> cdown:
the only reason you're running out of swap space is that you
have less memory than your applications want.
1483[17:18:04] <cdown> ratrace: These applications don't
want that memory. Their malloc decided they should get it as part of
the pool.
1484[17:18:18] <cdown> And that's often far unrelated to
what the developer actually intended.
1485[17:18:23] <ratrace> cdown: by "want" I mean
actively used pages
1486[17:18:36] <ratrace> I don't mean malloc 1PB on a 16GB
system and use maybe 1GB of it
1487[17:18:55] <jelly> ratrace, then don't look at Gnome app
VSS
1488[17:19:08] <ratrace> =)
1489[17:19:12] <jelly> they come close to that
1490[17:19:47] <genr8_> just say Noo to GNome
1491[17:19:50] <cdown> ratrace: It's simply not true,
I'm afraid. Using working set shrinking, we've reduced
physical memory workingset by huge amounts with no penalty to memory
pressure.
1492[17:20:03] <genr8_> NO*
1493[17:20:22] <ratrace> cdown: I think we're talking about
different things here
1495[17:21:16] <cdown> When I said "if swap gets full, you
may need to rethink your swap amount", your reply was simply
"no". I'm pretty sure we're talking about the
same thing.
1496[17:21:38] <jelly> cdown, if you fix it I will have to buy
you a beer, this was annoyed me for a decade or two
1499[17:22:09] <ratrace> bottom line of what I'm saying is,
if you have swapstorms then the simple asnwer is your applications
want more active memory at any given moment than your physical RAM
allows so the kernel is busy paging out what it thinks is inactive
and paging back in what's faulted on
1500[17:22:22] <cdown> jelly: My swap rework patchset to fix swap
behaviour hopefully will be out this half. I just need to make swap
shrinking work.
1501[17:22:34] <cdown> I'll let you know when it's
posted. I'm hoping around 5.11/5.12.
1502[17:23:30] <cdown> ratrace: Well, you mentioned swap storms a
lot later than your original comment ;-)
1503[17:23:30] <ratrace> but anyway, in context of that, the
larger the swap, the more pain there will be inflicted in such
swapstorm situations. ideally you want only a bit of swap, enough
RAM to accomodate active pages at any given moment, and don't
overcommit unless you really know what you're doing and that
most of overcommitted ram will be inactive and can be paged out in
vast amounts.
1504[17:23:45] <jelly> sometimes there are other weird reasons,
like NUMA making kernel decide it's faster to swap out than to
access RAM connected to the other cpu socket
1505[17:24:10] <jelly> suddently your database fills up only half
the RAM on a 2 socket machine, and the rest goes to swap
1507[17:24:30] <cdown> ratrace: Thrashing is tangential to the
page recycling amount, not the swap amount.
1508[17:24:37] <ratrace> cdown: well I'm a simple sysadmin,
not a kernel dev, so it's possible I used slightly.....
inappropriate or incomplete terminology. :)
1509[17:24:42] <cdown> s/tangential/proportional/
1510[17:24:59] <genr8_> what about this
replaced-url
1511[17:25:19] <genr8_> The idle page tracking feature allows to
track which memory pages are being accessed by a workload and which
are idle. This information can be useful for estimating the
workload’s working set size,
1512[17:25:33] <ratrace> cdown: yes but the more swap there is,
then more stuff can be paged out and if all that is suddenly
reclaimed.... :)
1513[17:25:36] <genr8_> The kernel internally keeps track of
accesses to user memory pages in order to reclaim unreferenced pages
first on memory shortage conditions
1561[17:35:04] <ratrace> under which circumstance would a file
change inode but nothing else (mtime, size, shasum, ownership,
xattrs)? or maybe that's somethign specific to ZFS (as
it's ZFS it happened on, already asked in #zfsonlinux, no
answer)
1562[17:35:57] <greycat> first thing that comes to mind: the file
was deleted, and then restored from a really comprehensive backup
1564[17:36:31] <genr8_> my first thought is the xattrs actually
were changed, and then changed back.
1565[17:36:53] <genr8_> wait no that wouldnt work
1566[17:37:00] <genr8_> ignore that
1567[17:37:17] <genr8_> greycat is right
1568[17:38:12] <ratrace> that didn't happen. I think
it's due to ZFS and what arcane magick it does in the
background when the fs is being snapshotted
1569[17:38:30] <dob1> greycat, I was testing the command before
using all the option I need for it... I don't deserve an ignore
for this
1570[17:38:47] *** Waggie2 is now known as Waggie
1571[17:39:13] <genr8_> for snapshots, it would be the same inode
unless it was written over again
1572[17:39:25] <genr8_> so something had to have been touched.
1573[17:39:34] <ratrace> or in other words, seems like it's
not a filesystem agnostic behavior unless it's what greycat
said, which isn't what happened, thus it must be zfs specific
1575[17:40:08] <genr8_> maybe you 'mv'ed it from one
dataset to another
1576[17:40:15] <ratrace> genr8_: now _that_ is possible yes.
it's roundcube temp files, and I have AIDE rules that exclude
file addition, removal, mtime changes, and today I saw an isolated
report that inode, and nothing else, changed on them
1577[17:40:28] *** Joins: metro (~metro@replaced-ip)
1601[17:52:34] <dpkg> <Aleric> pumpkin0: I noticed that
there are a LOT of capacitor on the the motherbord of 1200 micro
Farad (6.3 V), that ALL have 'burst open' ;) .. stuff is
coming out of it. I now think that it might be a reason for at least
extra instability.
1698[19:00:39] <greycat> that's certainly not *helping* the
benchmark
1699[19:00:42] <jelly> and it works!
1700[19:00:45] <neilthereildeil> i couldnt find dev/null...
1701[19:00:52] <greycat> you WHAT
1702[19:00:58] <ratrace> and on that note, I noticed /dev/urandom
-> /dev/null is slower than I'd expect it. I could swear I
saw it in hundreds of MB/s in previous years, now it's only 80
MB/s on the same computer with 4.19
1717[19:05:19] <ratrace> neilthereildeil: direct IO skips
filesystem pagecache; that said, I don't know if ramdisks are
subject to pagecache'ing at all
1765[19:37:20] <dpkg> Zen is :)(:, or a school of Mahayana
Buddhism, or the Zen kernel: a variant of the Linux kernel optimized
for desktop systems, including code not present in mainline.
Liquorix kernel image packages (replaced-url
1766[19:37:58] <TuxCrazy> jelly, thanks.
1767[19:38:13] <jelly> de nada
1768[19:38:41] <jelly> often the bot knows more than people do
1882[21:29:55] <slanck> does anyone know how to get the virtual
address from a struct page on Debian 8? I have been using
page_to_virt() that works fine with CentOS but the same operations
don't work on Debian 8
1883[21:30:01] *** Quits: earthundead (~earthunde@replaced-ip) (Remote host closed the connection)
1886[21:30:54] <slanck> I am trying to do that manually by using
gdb: (((unsigned long)(0xffffea0001d31200 - 0xffffea0000000000)
>> 6) << 12) + 0xffff880000000000
1887[21:31:10] <slanck> the command above work on CentOS 7 (3.10)
but doesn't work on Debian 8 (3.16)
1888[21:32:55] *** Quits: JohnML (~john1@replaced-ip) (Remote host closed the connection)
1889[21:35:10] <nvz> I'm not much of a programmer but that
sounds like a kernel thing to me, if not a glibc thing
1890[21:35:20] *** BenderRodriguez is now known as help
1945[22:31:24] *** Quits: deltanedas (~deltaneda@replaced-ip) (Remote host closed the connection)
1946[22:31:25] <jelly> GNU\colossus, I think there were some
separate images for the old macs requiring 32bit EFI, but it's
possible those were added to default installer lately
1947[22:31:31] <ratrace> firefox-esr-78 landed in sid. yay... and
stuff.
2085[23:38:35] <BadPractice> i try to get my pocket book (pb632)
to to sync with calibre but calibre does not see it. however it gets
mounted like a flash drive
2110[23:50:21] <jelly> ornxka, which debian release is this
supposed to be?
2111[23:50:36] <ornxka> its been so long i dont remember
2112[23:50:43] *** Quits: yuta (~pi@replaced-ip) (Quit: WeeChat 2.9)
2113[23:50:49] <jelly> if you're on sid, ask in
2114[23:50:51] <jelly> !debian-next
2115[23:50:52] <dpkg> #debian-next is the channel for
testing/unstable support on the OFTC network (irc.oftc.net), *not*
on freenode. If you get "Cannot join #debian-next (Channel is
invite only)." it means you did not read it's on
irc.oftc.net. See also
replaced-url
2121[23:51:12] <dpkg> [Basic Apt* Troubleshooting]. To diagnose
your problem, we need ALL OF THE FOLLOWING information: 1. complete
output of your apt-get/apt/aptitude run (including the command used)
2. output from "apt-cache policy pkg1 pkg2..." for ALL
packages mentioned ANYWHERE in the problem, and 3. "apt-cache
policy". Use
replaced-url
2123[23:51:31] <jelly> ornxka, what "apt-cache policy"
says is more useful.
2124[23:51:44] <ornxka> ah gotch
2125[23:51:45] <ornxka> a
2126[23:52:25] <jhutchins> !debian-next
2127[23:52:25] <dpkg> #debian-next is the channel for
testing/unstable support on the OFTC network (irc.oftc.net), *not*
on freenode. If you get "Cannot join #debian-next (Channel is
invite only)." it means you did not read it's on
irc.oftc.net. See also
replaced-url
2137[23:54:26] <jhutchins> jelly seems willing and able to help
you, and the information he requested will probably be needed in
-next too.
2138[23:54:58] <jhutchins> ornxka: It's pretty much expected
that things like that will happen in sid, and the only solution is
often just to wait it out.
2141[23:56:32] <jelly> well if you're on sid you're
expected to know or be able to learn your way around and maybe fix
things in ways that would not be recommended on Debian stable
2142[23:56:59] <jelly> like cherrypicking a snapshot from three
days ago and downgrading a couple packages to make things work
2146[23:58:26] <akp55> were can i go to learn about user mappings
for containers? i am using podman under root for running pods, and
mapping directories to container from my user to them, the directory
owner is getting changed. i'd like to understand how to prevent
that
2147[23:58:32] <adriangrigore> Hi, does anybody have idea what a
"*" referer is in an httpd log file?