| View previous topic :: View next topic |
| Author |
Message |
tld
Veteran
Joined: 09 Dec 2003
Posts: 1816
|
 Posted: Tue Nov 18, 2014 5:18 pm Post subject: |
 |
|
| saellaven wrote: | | Such actions should be configurable on a per demon basis... generally speaking, if a demon dies, I don't want it automatically restarted, I want it to stay down and notify me so that I can properly triage what is going on. Some people might want the demon automatically restarted, but that is a massive security and/or data integrity risk to assume that should be the default behavior for all or most services. |
That's been my first though every time I've seen this topic come up. I actually can't think of any situation where I'd want this ever. That is, in any case where a service is important enough that you need to keep it running, it's more important to know why it crashed, and blindly restarting is the last thing I'd want. |
|
| Back to top |
|
 |
depontius
Advocate
Joined: 05 May 2004
Posts: 3509
|
| Posted: Tue Nov 18, 2014 5:24 pm Post subject: |
|
|
| saellaven wrote: | | depontius wrote: |
3 - PID1 sits and waits for death signals from parentless processes, then passes the information to the second process in item #1 above.
4 - PID1 can restart the second process in item #1 above, if it dies. Some sort of backup action might be needed if it keeps dying too fast/easily.
Any "smart actions" upon process death are taken by the second process of item #1, perhaps using some sort of registered "helper function", which could be a daemon manager that has often been talked about. At the minimum is should probably make sure the pidfile states are correct. I don't even know if any changes to OpenRC to accomodate the above would be necessary.
|
Such actions should be configurable on a per demon basis... generally speaking, if a demon dies, I don't want it automatically restarted, I want it to stay down and notify me so that I can properly triage what is going on. Some people might want the demon automatically restarted, but that is a massive security and/or data integrity risk to assume that should be the default behavior for all or most services. |
Which is why I suggested it the way I did. It's rather kernel-like, in that the init system I propose doesn't enforce policy, it gives mechanisms for policy. You and the next commenter didn't want any auto-restart. I can see some people wanting some sort of restart. I can also see restarting some deamons, and not others. With any sort of auto-restart, I can see some sort of rate/count limiting, to prevent thrashing on a daemon that for some reason can't start.
_________________
.sigs waste space and bandwidth |
|
| Back to top |
|
|
Ant P.
Watchman
Joined: 18 Apr 2009
Posts: 6920
|
| Posted: Tue Nov 18, 2014 8:32 pm Post subject: |
|
|
It's trivial to do respawn rate-limiting on the likes of runit FWIW, you'd just put a "sleep" in the <service>/finish script. (if you wanted to be really sloppy about it, you could make that a symlink to /bin/sleep and it'd work...)
If you want to diagnose why it exited, you'd just make it sleep a few hours, to give your nagios or whatever monitor time to detect the failure, notify you of it, and give you time to sort it out manually. |
|
| Back to top |
|
|
ct85711
Veteran
Joined: 27 Sep 2005
Posts: 1791
|
| Posted: Tue Nov 18, 2014 8:37 pm Post subject: |
|
|
| Personally, I am with tld, in that I'd rather a service not be restarted, so I know there's an issue with it. I've already encountered one service, cupsd, that crashed on me a couple times, without any kind of logs/error messages outputted. Hell, if it's a critical service, it's even more important to know it's crashing as there's something wrong. If the admin is that lazy that they won't try figure out what's wrong and get talking with the developers to fix it, then that must not be all that critical. |
|
| Back to top |
|
|
Ottre
Tux's lil' helper
Joined: 23 Dec 2012
Posts: 129
|
| Posted: Tue Nov 18, 2014 8:40 pm Post subject: |
|
|
| saellaven wrote: | | Such actions should be configurable on a per demon basis... generally speaking, if a demon dies, I don't want it automatically restarted, I want it to stay down and notify me so that I can properly triage what is going on. |
Sounds like you'll love the next version of OpenRC, which features a plugin-based system for process supervision. :)
You add two lines to the configuration file in /etc/conf.d to turn it on, I believe it's off by default.
If you want to be notified when a daemon crashes, you can swap runit for monit, which includes email reports. |
|
| Back to top |
|
|
avx
Advocate
Joined: 21 Jun 2004
Posts: 2152
|
| Posted: Wed Nov 19, 2014 1:09 am Post subject: |
|
|
https://vote.debian.org/~secretary/gr_initcoupling/results.txt
Dingdingding, votes are in, fasten your seatbelts, there might be a wave of drama incoming.
| Quote: | The winners are:
Option 4 "General Resolution is not required" |
_________________
++++++++++[>+++++++>++++++++++>+++>+<<<<-]>++.>+.+++++++..+++.>++.<<+++++++++++++++.>.+++.------.--------.>+.>. |
|
| Back to top |
|
|
fourchannel
n00b
Joined: 21 Jul 2006
Posts: 17
|
| Posted: Wed Nov 19, 2014 6:32 am Post subject: |
|
|
| avx wrote: | https://vote.debian.org/~secretary/gr_initcoupling/results.txt
Dingdingding, votes are in, fasten your seatbelts, there might be a wave of drama incoming.
| Quote: | The winners are:
Option 4 "General Resolution is not required" |
|
That is probably going to set the debian fork group into action.
And I suspect that a sizable silent chunk of debian users is now seriously considering jumping ship.
_________________
---FourChannel---
Put your spare cpu cycles to good use |
|
| Back to top |
|
|
roki942
Apprentice
Joined: 18 Apr 2005
Posts: 285
Location: Seattle
|
| Posted: Wed Nov 19, 2014 6:52 am Post subject: |
|
|
| avx wrote: | https://vote.debian.org/~secretary/gr_initcoupling/results.txt
Dingdingding, votes are in, fasten your seatbelts, there might be a wave of drama incoming.
| Quote: | The winners are:
Option 4 "General Resolution is not required" |
|
Would some one be kind enough to explain what that all means? (Gentoo was the 1st distro I manged to get working on a computer and I have no real experience with others.) |
|
| Back to top |
|
|
djdunn
l33t
Joined: 26 Dec 2004
Posts: 810
|
| Posted: Wed Nov 19, 2014 7:34 am Post subject: |
|
|
i think it means nothing happens
_________________
“Music is a moral law. It gives a soul to the Universe, wings to the mind, flight to the imagination, a charm to sadness, gaiety and life to everything. It is the essence of order, and leads to all that is good and just and beautiful.”
― Plato |
|
| Back to top |
|
|
RazielFMX
l33t
Joined: 23 Apr 2005
Posts: 835
Location: NY, USA
|
| Posted: Wed Nov 19, 2014 1:44 pm Post subject: |
|
|
| roki942 wrote: | | avx wrote: | https://vote.debian.org/~secretary/gr_initcoupling/results.txt
Dingdingding, votes are in, fasten your seatbelts, there might be a wave of drama incoming.
| Quote: | The winners are:
Option 4 "General Resolution is not required" |
|
Would some one be kind enough to explain what that all means? (Gentoo was the 1st distro I manged to get working on a computer and I have no real experience with others.) |
I believe it means that the systemd crowd is free to do what they want to Debian and Debian is going to bend over and take it.
_________________
I am not anti-systemd; I am pro-choice. If being the latter makes you feel that I am the former, then so be it. |
|
| Back to top |
|
|
virtguru
Tux's lil' helper
Joined: 14 Aug 2010
Posts: 148
Location: The Greatest Country in the World
|
|
| Back to top |
|
|
avx
Advocate
Joined: 21 Jun 2004
Posts: 2152
|
| Posted: Wed Nov 19, 2014 2:08 pm Post subject: |
|
|
It means that Debian developers are free to chose if they want to support alternative init systems and that it is allowed for software to depend on a specific init.
--
Ian Jackson resigns from the TC, that's number #3 (on the TC alone) so far if I counted correctly.
https://lists.debian.org/debian-ctte/2014/11/msg00091.html
_________________
++++++++++[>+++++++>++++++++++>+++>+<<<<-]>++.>+.+++++++..+++.>++.<<+++++++++++++++.>.+++.------.--------.>+.>. |
|
| Back to top |
|
|
gwr
Apprentice
Joined: 19 Nov 2014
Posts: 194
|
| Posted: Wed Nov 19, 2014 4:12 pm Post subject: |
|
|
| It appears that systemd advocates are winning the war of attrition. |
|
| Back to top |
|
|
RazielFMX
l33t
Joined: 23 Apr 2005
Posts: 835
Location: NY, USA
|
| Posted: Wed Nov 19, 2014 5:12 pm Post subject: |
|
|
I'm starting to get tired of the pro-systemd camp making arguments to the non-pro-systemd camp (which encompasses the anti-systemd camp) that they should develop alternatives (such as systemd-shim).
Why do we need to build alternatives to something that many don't want or need in the first place? I feel like these discussions are all ushered into the realm of assuming that systemd is technically required/correct. You know what happens when you assume...
_________________
I am not anti-systemd; I am pro-choice. If being the latter makes you feel that I am the former, then so be it. |
|
| Back to top |
|
|
gwr
Apprentice
Joined: 19 Nov 2014
Posts: 194
|
| Posted: Wed Nov 19, 2014 8:25 pm Post subject: |
|
|
| RazielFMX wrote: | | I'm starting to get tired of the pro-systemd camp making arguments to the non-pro-systemd camp (which encompasses the anti-systemd camp) that they should develop alternatives (such as systemd-shim). |
It's worse than that. If I have existing applications to put behind the shims in order to emulate systemd, then I didn't need systemd in the first place. The entire project is a self-fullfilling tautology. |
|
| Back to top |
|
|
fourchannel
n00b
Joined: 21 Jul 2006
Posts: 17
|
| Posted: Wed Nov 19, 2014 8:31 pm Post subject: |
|
|
| Quote: | | It's worse than that. If I have existing applications to put behind the shims in order to emulate systemd, then I didn't need systemd in the first place. The entire project is a self-fullfilling tautology. |
That is precisely what I gathered from watching what "technologies" systemd was trying to implement. They didn't really make anything better, they just kept coming up with different and incompatible ways of doing things that were possible before systemd arrived and fucked everything up.
It's like they're not climbing some ladder towards new heights, but simply dragging everything in some sideways direction, and over a cliff it now looks.
To me, the entire systemd project / idea (?) arose from a lack of awareness of what can already be done in linux. It's like when I first learned to program on my graphing calculator, I was so excited that I kept adding more, and more, and more stuff for it to do. Until I was not able to understand how my program worked, and was completely unable to fix bugs when it misbehaved. It was too complicated. I would track down what I thought was causing the bug, change it, and break loads more stuff from working. And this problem arose precisely from my lack of knowledge of how to competently engineer programs. Now, as in years later, I have a degree in engineering, but I see the systemd crowd largely as inexperienced in managing complexity. It's like they have yet to learn the hard lesson in trying to fix hugely complicated software that was hacked together line by line with no design to guide them. I see their excitement for wanting everything to gather around systemd. I recognize it exactly as the same excitement I had when I first figured out how to program, and with none of the hard lessons I learned after programming terrible code.
systemd will fail under its own complexity, of that I have no doubt. But saving all the good parts of linux from being dragged under while systemd dies a slow death is now the battle.
Oh, and one more thing that I find irritating. systemd people keep saying how we have to keep up with new technology, so we must discard some of the old technology... Everyone on the planet loves GPS (I asked everyone : D ). A major part of GPS (not all) is using trigonometry to determine position. Trigonometry partly comes from the Pythagorean Theorem.
Do you know how old the Pythagoream Theorem is? Thousands of years old. But GPS fucking works.
Seriously people, just because time is flowing doesn't mean we need to discard old but 100% usable things. We build on old technology to make it better, only if it doesn't work do we scrap it (like..... alchemy).
Ok, rant over now.
_________________
---FourChannel---
Put your spare cpu cycles to good use |
|
| Back to top |
|
|
virtguru
Tux's lil' helper
Joined: 14 Aug 2010
Posts: 148
Location: The Greatest Country in the World
|
| Posted: Wed Nov 19, 2014 10:53 pm Post subject: |
|
|
| RazielFMX wrote: | I'm starting to get tired of the pro-systemd camp making arguments to the non-pro-systemd camp (which encompasses the anti-systemd camp) that they should develop alternatives (such as systemd-shim).
Why do we need to build alternatives to something that many don't want or need in the first place? I feel like these discussions are all ushered into the realm of assuming that systemd is technically required/correct. You know what happens when you assume... |
yeah well talk to LP about that. I am unprivileged of seeing him at a ccc congress some years back, what an ass. Did nothing but speak his crappy German/fake British accent and puke. Guess what , the masses loved it... He is articulate ONLY for his projects, but wont debate, hes an asshole heading a monstrosity of shit |
|
| Back to top |
|
|
mayak
n00b
Joined: 16 Jul 2013
Posts: 26
|
| Posted: Wed Nov 19, 2014 11:40 pm Post subject: |
|
|
| tr0ll wrote: | | yeah well talk to LP about that. I am unprivileged of seeing him at a ccc congress some years back, what an ass. Did nothing but speak his crappy German/fake British accent and puke. Guess what , the masses loved it... He is articulate ONLY for his projects, but wont debate, hes an asshole heading a monstrosity of shit |
please do not trash this thread - try to stick to technical arguments. otherwise they have close this thread.
| Anon-E-moose wrote: | Not talking about mom and pop shops or maybe 20 people corps, but large ones.
The buy in won't simply be about how smooth the transition is, but long term stability. |
I used to work for a large ISP with many thousand Linux systems. Their whole infrastructure was/is built around Red Hat. I don't see them migrating to any other OS because of sYstemd. Plus there is not much choice for commercial applications requiring 'enterprise linux'. SUSE, Oracle Linux and even Ubuntu use syStemd (or are going to). I hope you are right though. |
|
| Back to top |
|
|
virtguru
Tux's lil' helper
Joined: 14 Aug 2010
Posts: 148
Location: The Greatest Country in the World
|
| Posted: Thu Nov 20, 2014 12:14 am Post subject: |
|
|
| mayak wrote: | | tr0ll wrote: | | yeah well talk to LP about that. I am unprivileged of seeing him at a ccc congress some years back, what an ass. Did nothing but speak his crappy German/fake British accent and puke. Guess what , the masses loved it... He is articulate ONLY for his projects, but wont debate, hes an asshole heading a monstrosity of shit |
please do not trash this thread - try to stick to technical arguments. otherwise they have close this thread. |
While I see your validity, why thrashing? On what ?. Why are we trying to silence each other ? I was just stating an observation that was perceived. |
|
| Back to top |
|
|
truekaiser
l33t
Joined: 05 Mar 2004
Posts: 801
|
| Posted: Thu Nov 20, 2014 1:17 am Post subject: |
|
|
I have been mostly lurking in the shadows, checking in every once in a while so I can have a heads up if Gentoo does something stupid like make SystemD default.
But I have had an idea after reading LP's little manifesto of fundamentally changing how linux works into what 'he' thinks it should.
SystemD is FOSS right? Why not fork it?
Name the fork something like SystemE, code freeze the assimilation directive and first focus on bug fixing.
Sell it, so to speak, as a safer and more secure version of systemD that distro's can just drop in wholesale as a replacement. Later turn the hardlinks to soft dependencies while depreciating bad features like the binary only logging.
Or if you want to be ambitious, recode the logger to also dump text only logs first and then bug fix before doing the rest.
Same advantage, as in you need this if you even want to think about running gnome. Less headache of dealing with devs who refuse to fix bugs.
Best of all you make LP irrelevant in the long run as the linux distributions will want a version that has less headaches while acting mostly the same way.
call it, operation; Anything you can code, we can code better. |
|
| Back to top |
|
|
The Doctor
Moderator
Joined: 27 Jul 2010
Posts: 2678
|
| Posted: Thu Nov 20, 2014 1:35 am Post subject: |
|
|
| Quote: | | SystemD is FOSS right? Why not fork it? |
For the simple reason that the concept itself is what is broken. Changing who is developing it really doesn't change that.
_________________
First things first, but not necessarily in that order.
Apologies if I take a while to respond. I'm currently working on the dematerialization circuit for my blue box. |
|
| Back to top |
|
|
229566
Tux's lil' helper
Joined: 16 Aug 2010
Posts: 127
|
| Posted: Thu Nov 20, 2014 1:58 am Post subject: |
|
|
| truekaiser wrote: | | Why not fork it? |
Well, there's uselessd. Time will tell if stripping the ever increasing amount of code just to get to the "good idea" bits is becoming too much of an effort.
| Quote: | | Or if you want to be ambitious, recode the logger to also dump text only logs first and then bug fix before doing the 0rest. |
That would be essentially replacing one bad solution with another bad solution. The only "good idea" part of journalctl is indexing log entries, but hey, that could've been done with post-hoc log analyzers that can add graphics, audio, whatever bells and whistles a user could possibly want. In fact, it is already being done.
And in case post-hoc is insufficient because index has to be made from data not logged, but available at logging time, well, then patch the logger. Write an add-on for the logger that writes a separate binary index file to the existing text logs. Make it optional so people who don't care about it don't have to use it. Improve existing tools, don't scrap them and proceed with badly designed "replacements" that are not needed. |
|
| Back to top |
|
|
fourchannel
n00b
Joined: 21 Jul 2006
Posts: 17
|
| Posted: Thu Nov 20, 2014 2:09 am Post subject: |
|
|
| truekaiser wrote: | | SystemD is FOSS right? Why not fork it? |
systemd is not something really to seek after.
It's kinda like if you changed the colors and shapes of every road sign in america to something different. Then flipped them all upside down. Then assigned them all different meanings than what they were before.
Then declared that now the road signs in america are modern, and the situation is now so much better.
That is what systemd is. Not an improvement. Not helpful. Massively confusing to those who already know how to drive, and all the new drivers think the upside down road signs are great.
_________________
---FourChannel---
Put your spare cpu cycles to good use |
|
| Back to top |
|
|
truekaiser
l33t
Joined: 05 Mar 2004
Posts: 801
|
| Posted: Thu Nov 20, 2014 2:24 am Post subject: |
|
|
| The Doctor wrote: | | Quote: | | SystemD is FOSS right? Why not fork it? |
For the simple reason that the concept itself is what is broken. Changing who is developing it really doesn't change that. |
Well the idea was to first take control away from LP, then steer it to a more sane and less broken piece of code.
Making an alternative would mean you have to fight red-hat's entire marketing system and team. Think of it as co-opting a movement to take control of the ship it's on and turn it in a different direction. |
|
| Back to top |
|
|
saellaven
l33t
Joined: 23 Jul 2006
Posts: 646
|
| Posted: Thu Nov 20, 2014 3:28 am Post subject: |
|
|
| truekaiser wrote: | | The Doctor wrote: | | Quote: | | SystemD is FOSS right? Why not fork it? |
For the simple reason that the concept itself is what is broken. Changing who is developing it really doesn't change that. |
Well the idea was to first take control away from LP, then steer it to a more sane and less broken piece of code.
Making an alternative would mean you have to fight red-hat's entire marketing system and team. Think of it as co-opting a movement to take control of the ship it's on and turn it in a different direction. |
That assumes that we want any part of what they're doing...
In their world, it's cool to chop a leg off... what you're doing is suggest we chop an arm off instead. Me, I'd rather keep all of my appendages, not follow someone else's poor decisions.
In other words, their very concept is broken... emulating broken still gives you broken and even then you're assuming that they won't continue to routinely break stuff precisely in order to make sure their competition never has feature parity nor works right with the other software from the same company. |
|
| Back to top |
|
|
|
|
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
|
|
Gentoo Chat
