Re: Linux: concerns over systemd [OT]
- Original Message - From: Miles Fidelman mfidel...@meetinghouse.net Hmm, now this one I wasn't aware of this tidbit here has made this thread worthwhile to me, as we work on developing some clustered 'things' for use here. CoreOS wasn't even on the 'look at this at some point in time' list before, but it is now. Thanks, Jay. Funny, and here my reaction is just the opposite - to remove CoreOS from my list of things to look at. I am happy and sad for you both. Cheers, -- jra -- Jay R. Ashworth Baylink j...@baylink.com Designer The Things I Think RFC 2100 Ashworth Associates http://www.bcp38.info 2000 Land Rover DII St Petersburg FL USA BCP38: Ask For It By Name! +1 727 647 1274
Re: Linux: concerns over systemd [OT]
- Original Message - From: Gregory Boyce gregory.bo...@gmail.com On Wed, Oct 22, 2014 at 5:17 PM, Jeffrey Ollie j...@ocjtech.us wrote: I think that Debian's plan to allow multiple init systems (irregardless of which one is default) is a bad plan. The non-default ones won't get any love - at some point they'll just stop working (or indeed, work at all). If they break then one of two things will happen: 1) Someone will fix it. 2) No one will fix it because no one cares. If no one cares, then it being broken doesn't matter. Killing off choice/alternatives just in case no one cares about them isn't especially helpful. 3) A lot of people who do care and either cannot afford to or are technically competent to fix it are screwed. Cheers, -- jra -- Jay R. Ashworth Baylink j...@baylink.com Designer The Things I Think RFC 2100 Ashworth Associates http://www.bcp38.info 2000 Land Rover DII St Petersburg FL USA BCP38: Ask For It By Name! +1 727 647 1274
Re: Linux: concerns over systemd [OT]
Original Message - From: Jeffrey Ollie j...@ocjtech.us On Wed, Oct 22, 2014 at 9:48 PM, Jimmy Hess mysi...@gmail.com wrote: On Wed, Oct 22, 2014 at 1:31 PM, Barry Shein b...@world.std.com wrote: And you whisk all that away with it's not really clear to me that 'reboots in seconds' is a think to be optimized False dilemma. [ snip ] 10 seconds from power on to user interface for desktops, will meaningfully improve the user experience, but not for servers. It's a false dilemma only if you're thinking about traditional physical servers. Consider: 1) What if you're spinning up several thousand Hadoop nodes on AWS or GCE so that you can do some sort of big data operation. 2) What if PewDiePie just mentioned one of your products in a video and you need to quickly scale up the number of backend servers to handle the load. I'm sure that there are many other scenarios that I could devise where a fast server boot time was important. I will stipulate this use case. I will counter with you wouldn't be running a real distro in that case anyway; you'd be running something super trimmed down, and possibly custom built, or based on something like CoreOS, that only does one job. Well. :-) Cheers, -- jra -- Jay R. Ashworth Baylink j...@baylink.com Designer The Things I Think RFC 2100 Ashworth Associates http://www.bcp38.info 2000 Land Rover DII St Petersburg FL USA BCP38: Ask For It By Name! +1 727 647 1274
Re: Linux: concerns over systemd [OT]
after watching this discussion for a while, i have decided that i am in favour of systemd. i encourage its development, and widespread adoption. it will hasten the demise of linux in the server enviroment, which can only be a good thing. if people really want to run their servers on the *nix equivilent of Windows/XP, i say let them go ahead. every day that i have to work with linux, is another day i spend holding my nose. --jim -- Jim Mercer Reptilian Research j...@reptiles.org+1 416 410-5633 He who dies with the most toys is nonetheless dead
Re: Linux: concerns over systemd [OT]
On Mon, Oct 27, 2014 at 10:35 AM, Jay Ashworth j...@baylink.com wrote: I will stipulate this use case. I will counter with you wouldn't be running a real distro in that case anyway; you'd be running something super trimmed down, and possibly custom built, or based on something like CoreOS, that only does one job. Well. :-) From: https://coreos.com/using-coreos/systemd/ CoreOS uses systemd as the core of its distributed init system, fleet. Systemd is well supported in many Linux distros, making it familiar to most engineers. Every aspect of CoreOS is deeply integrated with systemd. -- Jeff Ollie
Re: Linux: concerns over systemd [OT]
- Original Message - From: Jeffrey Ollie j...@ocjtech.us On Mon, Oct 27, 2014 at 10:35 AM, Jay Ashworth j...@baylink.com wrote: I will stipulate this use case. I will counter with you wouldn't be running a real distro in that case anyway; you'd be running something super trimmed down, and possibly custom built, or based on something like CoreOS, that only does one job. Well. :-) From: https://coreos.com/using-coreos/systemd/ CoreOS uses systemd as the core of its distributed init system, fleet. Systemd is well supported in many Linux distros, making it familiar to most engineers. Every aspect of CoreOS is deeply integrated with systemd. Surprisingly, I actually knew this already. You might want to stop trying to score points, rather than actually, y'know, just advancing the conversation. Cheers, -- jra -- Jay R. Ashworth Baylink j...@baylink.com Designer The Things I Think RFC 2100 Ashworth Associates http://www.bcp38.info 2000 Land Rover DII St Petersburg FL USA BCP38: Ask For It By Name! +1 727 647 1274
Re: Linux: concerns over systemd [OT]
On 10/27/2014 11:35 AM, Jay Ashworth wrote: I will counter with you wouldn't be running a real distro in that case anyway; you'd be running something super trimmed down, and possibly custom built, or based on something like CoreOS, that only does one job. Well. Hmm, now this one I wasn't aware of this tidbit here has made this thread worthwhile to me, as we work on developing some clustered 'things' for use here. CoreOS wasn't even on the 'look at this at some point in time' list before, but it is now. Thanks, Jay.
Re: Linux: concerns over systemd [OT]
Lamar Owen wrote: On 10/27/2014 11:35 AM, Jay Ashworth wrote: I will counter with you wouldn't be running a real distro in that case anyway; you'd be running something super trimmed down, and possibly custom built, or based on something like CoreOS, that only does one job. Well. Hmm, now this one I wasn't aware of this tidbit here has made this thread worthwhile to me, as we work on developing some clustered 'things' for use here. CoreOS wasn't even on the 'look at this at some point in time' list before, but it is now. Thanks, Jay. Funny, and here my reaction is just the opposite - to remove CoreOS from my list of things to look at. Cheers, Miles Fidelman
Re: Linux: concerns over systemd [OT]
On Thu, Oct 23, 2014 at 12:28 AM, George Herbert george.herb...@gmail.com wrote: Ok. As a highly on- list-topic example of why distrust is called for... Without referring to the systemd source code*, does anyone know what systemd uses to select between networking subsystems (i.e. NetworkManager, the new standard as of RHEL 7, vs /etc/ sysconfig/network-scripts/, etc.). NetworkManager is default but disableable and it magically falls back to network-scripts dir, but the fallback is nearly undocumented and the selection behavior appears completely undocumented. systemctl status NetworkManager.service systemctl status network.service I don't think that there's anything magic about it, you have one or the other enabled. Adding NM_CONTROLLED=yes/no to /etc/sysconfig/network-scripts/ifcfg-* gives you per-interface control over whether NetworkManager or the network scripts are used for managing the interface. If neither is enabled you probably end up with no networking. If by some chance you do know this, where did you come by that knowledge? Hopefully with URLs. I have access to systems that run systemd and I tried a couple of things... Also, I've been managing Red Hat systems for a long time and have known about this for a while. But a little bit of googling and I found this: https://access.redhat.com/documentation/en-US/Red_Hat_Enterprise_Linux/7/html/Networking_Guide/sec-NetworkManager_and_the_Network_Scripts.html Unless you're running systemd-networkd, this is really distro-specific stuff as I expect that most distros will want to preserve some backward compatibility with legacy network configuration. -- Jeff Ollie
Re: Linux: concerns over systemd [OT]
When I'm talking about hardware initialization, I'm talking about the huge part that appends *before* the kernel boots. For example, hard-based RAID. On my server, when I push the start button, bios start-up, do a lot of awesome things (irony), start the raid (sloowly), and then, after 5min, pop up grub, then Linux, etc. 5min for hard (from power-on to grub), 20sec for software (from grub to prompt)
Re: Linux: concerns over systemd [OT]
On Tue, Oct 21, 2014 at 11:59:44PM +0100, Tom Hill wrote: It's Gentoo: You should write your own is the most likely answer. Not if you ask nicely :) -- Eray Aslan e...@gentoo.org
Re: Linux: concerns over systemd [OT]
On Wed, Oct 22, 2014 at 5:28 PM, Gregory Boyce gregory.bo...@gmail.com wrote: On Wed, Oct 22, 2014 at 5:17 PM, Jeffrey Ollie j...@ocjtech.us wrote: I think that Debian's plan to allow multiple init systems (irregardless of which one is default) is a bad plan. The non-default ones won't get any love - at some point they'll just stop working (or indeed, work at all). If they break then one of two things will happen: 1) Someone will fix it. 2) No one will fix it because no one cares. If no one cares, then it being broken doesn't matter. Killing off choice/alternatives just in case no one cares about them isn't especially helpful. Resending since my subscription was apparently a bit off. -- Greg
Re: Linux: concerns over systemd [OT]
On October 22, 2014 at 15:31 jfb...@gmail.com (Ricky Beam) wrote: On Wed, 22 Oct 2014 14:31:02 -0400, Barry Shein b...@world.std.com wrote: Perhaps you don't remember the days when an fsck was basically mandatory and could take 15-20 minutes on a large disk. Journaling has all but done away with fsck. You'd have to go *way* back to have systems that ran a full fsck on every boot -- and in my experience, you absolutely wanted that fsck. That was my point, it was a very brief and concise 30 year history. That's why I mentioned the introduction of the clean bit which was when we began recording (there may have been earlier experiments) the clean unmounting of a file system in the superblock so no need to fsck. And you whisk all that away with it's not really clear to me that 'reboots in seconds' is a think to be optimized (I hope it's clear I meant thing to be opt...) Your efforts are better spent avoiding an outage in the first place. If outages are common enough to be something that needs to be sped up, then you've already failed. One important tool is failover. But once a system fails over you'd like to see the failed component back in service as quickly as possible unless you have an infinite number of redundant systems. Your advice doesn't ring true to me. -- -Barry Shein The World | b...@theworld.com | http://www.TheWorld.com Purveyors to the Trade | Voice: 800-THE-WRLD| Dial-Up: US, PR, Canada Software Tool Die| Public Access Internet | SINCE 1989 *oo*
Re: Linux: concerns over systemd [OT]
On October 23, 2014 at 04:42 ra...@psg.com (Randy Bush) wrote: Barry Schein: Interesting you went to the trouble to add a 'c' to my name! You need better quoting tools. I'm reminded of the remark often attributed to DEC CEO Ken Olson, roughly: With VMS (their big complex OS) it might take hours searching through manuals to find a feature you need while with Unix you can determine in seconds that it is not available. and how did that work out for vms? and digital? A few people made billions, a few more made many millions, hundreds of thousands (or thereabouts) had pretty good jobs for upwards of 20 years, and then the second largest computer company in the world vaporized almost mysteriously. The VAX hardware was important. It was for the time relatively inexpensive and very capable, the 32-bit address space (ok, technically four 30 bit addr spaces) and VM hardware at those prices were revolutionary. You had many of the capabilities of a multi-million dollar mainframe for about 1/10th the cost. Ran Unix great! VMS not so much. Mostly re-warmed over RSX (an earlier DEC OS) with a few new ideas to take advantage of the platform, and some cobbling from their TOPS-10 and TOPS-20 OS's (e.g. galaxy.) IMHO DEC desparately wanted to go head on with IBM's 370 line but just didn't seem to get why companies bought IBM mainframes, or found those parts too expensive to compete on. But they did ok financially anyhow so who's to criticize? VMS even had PIP! And sometimes you needed it. -- -Barry Shein The World | b...@theworld.com | http://www.TheWorld.com Purveyors to the Trade | Voice: 800-THE-WRLD| Dial-Up: US, PR, Canada Software Tool Die| Public Access Internet | SINCE 1989 *oo*
Re: Linux: concerns over systemd [OT]
Going way off topic but what's still a disaster in log files is the lack of standardization of output. As another extreme OS/370 catalogued virtually (hah) every error msg, if you thought you had a new one you added it to the catalogue as you added it to an error msg in your program and it was likely someone informed you something sufficient already existed, or you just specialized an existing code -- e.g., IED101203EA77... might mean daemon, file system problem, insufficient privilege, recoverable/unrecoverable, etc and then you could add a few more digits (...) to make it unique if you liked or use a known value and some free format text as per usual if desired. System/Kernel/Library wide. I realize there have been a few very weak attempts at this with *ix like errno, strerror (which for some bizarre reason never prints the errno or symbolic error only some text albeit from a known table), sysexits.h, %m in syslog which is just strerror(), etc. But syslog et al needs to go way beyond the daemon, time, and priority and free format text so log analyzers (including grep) have half a chance. Just my 2c. -- -Barry Shein The World | b...@theworld.com | http://www.TheWorld.com Purveyors to the Trade | Voice: 800-THE-WRLD| Dial-Up: US, PR, Canada Software Tool Die| Public Access Internet | SINCE 1989 *oo*
Re: Linux: concerns over systemd [OT]
On Oct 21, 2014, at 6:03 PM, Jay Ashworth j...@baylink.com wrote: GNOME is probably the linchpin. But it's not just RH. It's Debian, and by extension *buntu, and SuSE, and at least one other major independent parent distro that I can't think of just now... And as far as I know, it's done; SuSE packages already largely don't even include initscripts. Enough to make a grown man fork RHEL (or, CentOS). George William Herbert Sent from my iPhone
Re: Linux: concerns over systemd [OT]
George Herbert wrote: On Oct 21, 2014, at 6:03 PM, Jay Ashworth j...@baylink.com wrote: GNOME is probably the linchpin. But it's not just RH. It's Debian, and by extension *buntu, and SuSE, and at least one other major independent parent distro that I can't think of just now... And as far as I know, it's done; SuSE packages already largely don't even include initscripts. Enough to make a grown man fork RHEL (or, CentOS). Which leads me to ask - those of you running server farms - what distros are popular these days, for server-side operations? We've been running Debian like forever (by way of Solaris and redhat) - but this systemd thing is making me rethink things. Seems like an awful lot of folks are now designing for the desktop, and it might be time to migrate to a BSD or Solaris derivative. What are others doing? Miles Fidelman -- In theory, there is no difference between theory and practice. In practice, there is. Yogi Berra
Re: Linux: concerns over systemd [OT]
On 22/10/14 10:41, Miles Fidelman wrote: Which leads me to ask - those of you running server farms - what distros are popular these days, for server-side operations? We've been running Debian like forever (by way of Solaris and redhat) - but this systemd thing is making me rethink things. Seems like an awful lot of folks are now designing for the desktop, and it might be time to migrate to a BSD or Solaris derivative. What are others doing? Not making rash decisions. Debian and CentOS are still the 'asked for' distributions of Linux. Once in a blue moon, someone asks about something else. Those that care are outnumbered greatly by those that just want a known platform to develop upon. (The irony of this is not lost on me). I'd take systemd over ditching apt/yum in a mad panic. And I'm certainly no fan of systemd myself. -- Tom
Re: Linux: concerns over systemd [OT]
Before leaving Debian, things to think: - will systemd be officialy the only system available ? - if so, won't we get a way to bypass that ? I'm not gonna throw Debian away due to such a mess, without fighting hard, and I think you should do the same: talk, patch if needed, show you're here If all good people who laugh insanely about systemd leave Debian alone, who's left ? gnome-people ? systemd-fanatics (heretics?) ? On 22/10/2014 12:09, Tom Hill wrote: On 22/10/14 10:41, Miles Fidelman wrote: Which leads me to ask - those of you running server farms - what distros are popular these days, for server-side operations? We've been running Debian like forever (by way of Solaris and redhat) - but this systemd thing is making me rethink things. Seems like an awful lot of folks are now designing for the desktop, and it might be time to migrate to a BSD or Solaris derivative. What are others doing? Not making rash decisions. Debian and CentOS are still the 'asked for' distributions of Linux. Once in a blue moon, someone asks about something else. Those that care are outnumbered greatly by those that just want a known platform to develop upon. (The irony of this is not lost on me). I'd take systemd over ditching apt/yum in a mad panic. And I'm certainly no fan of systemd myself.
Re: Linux: concerns over systemd [OT]
na...@jack.fr.eu.org (na...@jack.fr.eu.org) wrote: I'm not gonna throw Debian away due to such a mess, without fighting hard, and I think you should do the same: talk, patch if needed, show you're here ...and sit it out with wheezy-LTS... Elmar.
Re: Linux: concerns over systemd [OT]
Which leads me to ask - those of you running server farms - what distros are popular these days, for server-side operations? been running bsd forever. but moving to debian and ganeti, as bsd does not host virtualization. would love it if debian ditched this systemd monstrosity and provided solid zfs. randy
Re: Linux: concerns over systemd [OT]
Which leads me to ask - those of you running server farms - what distros are popular these days, for server-side operations? been running bsd forever. but moving to debian and ganeti, as bsd does not host virtualization. Simply not true; http://bhyve.org/ It is a bit immature compared to Xen+Ganeti or something like that. would love it if debian ditched this systemd monstrosity and provided solid zfs. ... JG -- Joe Greco - sol.net Network Services - Milwaukee, WI - http://www.sol.net We call it the 'one bite at the apple' rule. Give me one chance [and] then I won't contact you again. - Direct Marketing Ass'n position on e-mail spam(CNN) With 24 million small businesses in the US alone, that's way too many apples.
Re: Linux: concerns over systemd [OT]
On 22 October 2014 11:34, na...@jack.fr.eu.org wrote: Before leaving Debian, things to think: - will systemd be officialy the only system available ? - if so, won't we get a way to bypass that ? And one other thought... is it really that bad? Personally I like it a lot better than sysV plus inittab plus daemontools. Dan
Re: Linux: concerns over systemd [OT]
Which leads me to ask - those of you running server farms - what distros are popular these days, for server-side operations? been running bsd forever. but moving to debian and ganeti, as bsd does not host virtualization. Simply not true; http://bhyve.org/ It is a bit immature compared to Xen+Ganeti or something like that. apologies. i thought we were talking about production systems. my mistake. randy
Re: Linux: concerns over systemd [OT]
Which leads me to ask - those of you running server farms - what distros are popular these days, for server-side operations? been running bsd forever. but moving to debian and ganeti, as bsd does not host virtualization. Simply not true; http://bhyve.org/ It is a bit immature compared to Xen+Ganeti or something like that. apologies. i thought we were talking about production systems. my mistake. Oh, c'mon Randy, you've been around long enough to know how this all works. You can't honestly tell me that VMware ESX was born handling production loads. You can't honestly tell me that Xen was born handling production loads. All hypervisor technologies were new at one point in their life cycle, and most were also catastrofails at one point in their life cycle. The fact that bhyve is new means it's more immature, but people are certainly trying noncricitical production loads on it. Y'know, the same way they did years ago with ESX. No one's saying you have to trust it with your production workloads, but it's pretty unfair to characterize BSD as not host(ing) virtualization when so much effort has been put into that very issue, specifically so that we could gain the advantages of a BSD hypervisor that supported ZFS natively... ... JG -- Joe Greco - sol.net Network Services - Milwaukee, WI - http://www.sol.net We call it the 'one bite at the apple' rule. Give me one chance [and] then I won't contact you again. - Direct Marketing Ass'n position on e-mail spam(CNN) With 24 million small businesses in the US alone, that's way too many apples.
Re: Linux: concerns over systemd [OT]
On Wed, Oct 22, 2014 at 12:00:52PM +0100, Daniel Ankers wrote: On 22 October 2014 11:34, na...@jack.fr.eu.org wrote: Before leaving Debian, things to think: - will systemd be officialy the only system available ? - if so, won't we get a way to bypass that ? And one other thought... is it really that bad? Personally I like it a lot better than sysV plus inittab plus daemontools. When it was init+daemontools, I could hold my nose over the binary logging and consider using it. Now it's taking over cron and all manner of other things, there's no way in hell I'm letting it onto my systems. As to the issue of will it only be systemd, the problem is that as the officially-blessed option, that's the one that'll get the universal support, so if you want to run something else, some things will mysteriously not work, and package maintainers won't care nearly as much. Bypassing systemd is a whole hell of a lot harder than switching out sysvinit for something else, because systemd does so many things, and many other things are being modified to absolutely depend on things that only systemd provides -- GNOME's the big one, but docker is closely tied to systemd too, I believe, I think udev needs systemd now (or has it been incorporated into systemd? I can't keep all this straight) and I'm pretty sure I've heard of other things deprecating non-systemd ways of doing things. The *really* damaging part of it, though, is that as systemd grows to overshadow the things it re-implements (udev, dbus, etc) it starves the alternatives of light and they quickly fall behind and are no longer viable as ways to avoid systemd. That isn't systemd's fault, per se, but it does make it much harder over time to avoid getting caught in the gaping maw. - Matt
Re: Linux: concerns over systemd [OT]
When it's working, no doupt, I'll be fine I don't care (or just a few) about when it's working. The point is: what about it's failure ? On the ethical point of view, systemd is killed anyway On 22/10/2014 13:00, Daniel Ankers wrote: On 22 October 2014 11:34, na...@jack.fr.eu.org mailto:na...@jack.fr.eu.org wrote: Before leaving Debian, things to think: - will systemd be officialy the only system available ? - if so, won't we get a way to bypass that ? And one other thought... is it really that bad? Personally I like it a lot better than sysV plus inittab plus daemontools. Dan
Re: Linux: concerns over systemd [OT]
On Wed, Oct 22, 2014 at 6:12 AM, Joe Greco jgr...@ns.sol.net wrote: when so much effort has been put into that very issue, specifically so that we could gain the advantages of a BSD hypervisor that supported ZFS natively... [snip] If you want native ZFS support, then Solaris x86-64+Zones+KVM or SmartOS. Now Solaris/Illumos' process supervision and fault management systems, SMF and FMA are pretty complex, but they aren't stuffing 1000 random tasks into the init program. ^_^ ... JG -- -JH
Re: Linux: concerns over systemd [OT]
On Wed, 22 Oct 2014 12:34:17 +0200, na...@jack.fr.eu.org said: Before leaving Debian, things to think: - will systemd be officialy the only system available ? - if so, won't we get a way to bypass that ? Somebody already forked systemd at a point before it completely lost the plot. http://uselessd.darknedgy.net/ pgpdhmr9dzrU5.pgp Description: PGP signature
Re: Linux: concerns over systemd [OT]
na...@jack.fr.eu.org wrote: Before leaving Debian, things to think: - will systemd be officialy the only system available ? - if so, won't we get a way to bypass that ? officially, there will be support for multiple init systems; in practice, the installer doesn't provide an option, and trying to remove systemd seems to drag one into dependency hell I'm not gonna throw Debian away due to such a mess, without fighting hard, and I think you should do the same: talk, patch if needed, show you're here unfortunately, the fight is rather ugly, and seemingly fruitless (have you been on debian-user, debian-devel, or debian-vote lately?) - sigh... this discussion, if a bit OT for nanog, is a breath of fresh air - PLEASE weigh in on the side of goodness and light Miles Fidelman -- In theory, there is no difference between theory and practice. In practice, there is. Yogi Berra
Re: Linux: concerns over systemd [OT]
I'm reminded of the remark often attributed to DEC CEO Ken Olson, roughly: With VMS (their big complex OS) it might take hours searching through manuals to find a feature you need while with Unix you can determine in seconds that it is not available. On October 21, 2014 at 16:10 asulli...@dyn.com (Andrew Sullivan) wrote: On Tue, Oct 21, 2014 at 03:11:55PM -0400, Barry Shein wrote: But for example some of my servers boot in seconds. One is reminded of a mail, included in the Preface to _The UNIX-HATERS Handbook_, available at http://www.art.net/~hopkins/Don/unix-haters/preface.html. Apparently, things really are going to get a lot worse before they get worse. Best regards, A -- Andrew Sullivan Dyn, Inc. asulli...@dyn.com v: +1 603 663 0448 -- -Barry Shein The World | b...@theworld.com | http://www.TheWorld.com Purveyors to the Trade | Voice: 800-THE-WRLD| Dial-Up: US, PR, Canada Software Tool Die| Public Access Internet | SINCE 1989 *oo*
Re: Linux: concerns over systemd [OT]
On October 21, 2014 at 16:43 morrowc.li...@gmail.com (Christopher Morrow) wrote: On Tue, Oct 21, 2014 at 4:10 PM, Andrew Sullivan asulli...@dyn.com wrote: On Tue, Oct 21, 2014 at 03:11:55PM -0400, Barry Shein wrote: But for example some of my servers boot in seconds. One is reminded of a mail, included in the Preface to _The UNIX-HATERS Handbook_, available at it's really not clear to me that 'reboots in seconds' is a thing to optimize... The unix community has exerted great amounts of effort over the decades to speed up reboot, particularly after crashes but also planned. Perhaps you don't remember the days when an fsck was basically mandatory and could take 15-20 minutes on a large disk. Then we added the clean bit (disk unmounted cleanly, no need for fsck), reorg'd the file system layout to speed up fsck considerably and make it more reliable/recoverable, added journaled file systems which really sped things up often eliminating the need to fsck after a crash entirely and recovering in seconds, various attempts to figure out the dependency graph of servers and services which need to be started so they could be started in parallel where dependencies are met, etc. And learned how to do hot failover and master/slave servers etc. And you whisk all that away with it's not really clear to me that 'reboots in seconds' is a think to be optimized To me that's like saying it's not important to try to design so one can recover from a network outage in seconds. Anyhow, if it's not clear: I disagree. I suppose the win is: Is the startup/shutdown process clear, conscise and understandable at 3am local time? followed by: Can I adjust my startup processes to meet my needs easily and without finding a phd in unix? If systemd is simply a change in how I think about /etc/init.d/* and /etc/rc?.d/* cool, if it's more complexity and less EASY flexibility then it's a fail. Actually, much of that is less important except perhaps to a hobbyist. You only have to get the startup/shutdown process etc right once in a while and generally during a planned outage. Recovering from a failure or going back into service quickly after a planned outage is critical and can be critical at any time. Obviously one can appeal to extremum but what you say doesn't make sense to me. At any rate, you are disputing a huge, decades long, and widely fought battle. It's certainly not my opinion. -- -Barry Shein The World | b...@theworld.com | http://www.TheWorld.com Purveyors to the Trade | Voice: 800-THE-WRLD| Dial-Up: US, PR, Canada Software Tool Die| Public Access Internet | SINCE 1989 *oo*
Re: Linux: concerns over systemd [OT]
On Wed, 22 Oct 2014 14:31:02 -0400, Barry Shein b...@world.std.com wrote: Perhaps you don't remember the days when an fsck was basically mandatory and could take 15-20 minutes on a large disk. Journaling has all but done away with fsck. You'd have to go *way* back to have systems that ran a full fsck on every boot -- and in my experience, you absolutely wanted that fsck. I've used xfs for over a decade. It doesn't even have an fsck (xfs_check and xfs_repair, yes, but NO system will ever call them. And as a rule, never needs to.) And you whisk all that away with it's not really clear to me that 'reboots in seconds' is a think to be optimized You're arguing the difference between optimizing a 15min boot into a 5min boot, vs a 15sec boot into a 14sec boot. The former is an actual optimization allowing subsystems to start in parallel, in ways that do not introduce delay. (eg. sendmail startup, used to be the #1 slow down on solaris) The latter is pure nonsense, with boot being measured as when login pops up -- which is NOT when all the subsystems have actually completely started. To me that's like saying it's not important to try to design so one can recover from a network outage in seconds. Your efforts are better spent avoiding an outage in the first place. If outages are common enough to be something that needs to be sped up, then you've already failed.
Re: Linux: concerns over systemd [OT]
Which leads me to ask - those of you running server farms - what distros are popular these days, for server-side operations? We've been running Debian like forever (by way of Solaris and redhat) - but this systemd thing is making me rethink things. Seems like an awful lot of folks are now designing for the desktop, and it might be time to migrate to a BSD or Solaris derivative. What are others doing? to be honest, i like systemd. nobody else has really stepped up to the bat to fix issues of existing init systems and tying interoperabilty into a common bus. not everything in systemd is a requirement to run it. just because a unit is offered for dhcp or ntp, doesn't mean you are required to use it. the vast majority of negative tongue wagging regarding systemd is ill informed. does systemd have growing pains? definitely. are some egos involved? sure. can systemd be far reaching? yes, is such reach mandated? no. use the units you want and disregard the rest. i run Arch Linux in my farms and except for the advanced networking functions and ease of locally tailoring Gentoo offered, I love Arch more than any of my .rpm or .deb flavors, *bsd or *nix, and i have used pretty much all of them. i've run unix like systems since 1988 and cover the gammut of code writing from bash scripts to kernel modules to apache and postgres on everything from microcontrollers to mainframes. before making ill informed rash decisions, study what you're talking about and then decide whether or not you like and can/can't use said subject. -d signature.asc Description: OpenPGP digital signature
Re: Linux: concerns over systemd [OT]
On October 22, 2014 at 12:00 md1...@md1clv.com (Daniel Ankers) wrote: On 22 October 2014 11:34, na...@jack.fr.eu.org wrote: Before leaving Debian, things to think: - will systemd be officialy the only system available ? - if so, won't we get a way to bypass that ? And one other thought... is it really that bad? Personally I like it a lot better than sysV plus inittab plus daemontools. I posted my complaints but I think they fall more in the realm of lack of maturity than bad design. I believe systemd is superior to sysvinit but it will take time for it to mature, administrative tools to become available (even if just better logging/tracing), and for us to get used to it and acquire the folk knowledge we need. Until then frustration will arise from time to time. -- -Barry Shein The World | b...@theworld.com | http://www.TheWorld.com Purveyors to the Trade | Voice: 800-THE-WRLD| Dial-Up: US, PR, Canada Software Tool Die| Public Access Internet | SINCE 1989 *oo*
Re: Linux: concerns over systemd [OT]
Barry Schein: I'm reminded of the remark often attributed to DEC CEO Ken Olson, roughly: With VMS (their big complex OS) it might take hours searching through manuals to find a feature you need while with Unix you can determine in seconds that it is not available. and how did that work out for vms? and digital? Jeffrey Ollie wrote: The people that like systemd (like myself) have wisely learned that the people that hate systemd, hate it mostly because it's different from what came before and don't want to change. There's no way to argue rationally with that. and with this ad homina you dismiss all technical, architectural, and security concerns. much from folk who have been administering unix systems since you were in nappies. and you advocate rational argument? Daniel Corbe co...@corbe.net wrote: Not to get even further off topic here but when was the last time you maintained a BSD system? FreeBSD (at least) adopted binary package management as its preferred interface to ports through pkg-ng somewhere in the 9-RELEASE cycle. i am a long time bsd user and deployer. amelioration of shellshock, heartbleed, and poodle on an annoying number of servers was not any better on freebsd than debian. randy
Re: Linux: concerns over systemd [OT]
David Ford wrote: Which leads me to ask - those of you running server farms - what distros are popular these days, for server-side operations? We've been running Debian like forever (by way of Solaris and redhat) - but this systemd thing is making me rethink things. Seems like an awful lot of folks are now designing for the desktop, and it might be time to migrate to a BSD or Solaris derivative. What are others doing? snip before making ill informed rash decisions, study what you're talking about and then decide whether or not you like and can/can't use said subject. Is that not the point of discussions like this one? -- In theory, there is no difference between theory and practice. In practice, there is. Yogi Berra
Re: Linux: concerns over systemd [OT]
Which leads me to ask - those of you running server farms - what distros are popular these days, for server-side operations? We've been running Debian like forever (by way of Solaris and redhat) - but this systemd thing is making me rethink things. Seems like an awful lot of folks are now designing for the desktop, and it might be time to migrate to a BSD or Solaris derivative. What are others doing? to be honest, i like systemd. nobody else has really stepped up to the bat to fix issues of existing init systems and tying interoperabilty into a common bus. Perhaps because folks that understand more about security than you (and me for sure so I'm not picking on you) think thats a bad idea? If something is a bad idea then smart folks dont rush out (generally) to build it ... thus the no one stepping up to bat problem thats not really a problem - its a good thing to not have problems solved improperly. Perhaps because when you say/hear things like tying interoperabilty into a common bus you think thats a good idea. Others hear those same words and think: vendor lock-in single point of failure lack of choice The binary logging thing is a non-starter for a lot of folks. dbus ? On a server ? Do we really need that ? Lets keep servers reliable - less code not more (no bugs in unwritten code). Shouldnt the amount of code running as PID 1 be kept to an absolute minimum? Bad architecture decisions dont suddenly become good ones even if they solve other problems along the way or make some things better or faster.
Re: Linux: concerns over systemd [OT]
the vast majority of negative tongue wagging regarding systemd is ill informed. can we skip the ad homina and leave that for the systemd dev fora? does systemd have growing pains? definitely. are some egos involved? sure. can systemd be far reaching? yes, is such reach mandated? no. use the units you want and disregard the rest. how does this work out in practice? at install, can i choose whether systemd is used for X as opposed for the separate component? can i template such choices for cluster deployment with the usual tools? as barry said, we get to drive this car. randy
Re: Linux: concerns over systemd [OT]
Randy Bush wrote: the vast majority of negative tongue wagging regarding systemd is ill informed. can we skip the ad homina and leave that for the systemd dev fora? does systemd have growing pains? definitely. are some egos involved? sure. can systemd be far reaching? yes, is such reach mandated? no. use the units you want and disregard the rest. how does this work out in practice? at install, can i choose whether systemd is used for X as opposed for the separate component? can i template such choices for cluster deployment with the usual tools? Right now, the problem is that the installer does not give a choice, and the preseed function has a bug such that you can't do a systemvinit install directly at install time; rather you have to let the default install complete, then replace systemd with sysvinit-core - and there seem to be cases of dropping into dependency hell when doing so (reports vary). Longer term, the above-mentioned bug will (hopefully) be patched (the patch is there, but not well tested, and has not yet been included into the installer). Then we get into the question of to what extent upstream packages start depending on systemd. Miles Fidelman -- In theory, there is no difference between theory and practice. In practice, there is. Yogi Berra
Re: Linux: concerns over systemd [OT]
On Wed, 22 Oct 2014 19:35:51 -, David Ford said: into a common bus. not everything in systemd is a requirement to run it. just because a unit is offered for dhcp or ntp, doesn't mean you are required to use it. Actually, systemd 216 will cram systemd-timesyncd down your throat even if you had ntpd installed. https://bugzilla.redhat.com/show_bug.cgi?id=1136905 https://mail.gnome.org/archives/distributor-list/2014-September/msg2.html Lennart's attitude was pretty much why would anybody want to run ntpd when they have our SNTP implementation: http://lists.freedesktop.org/archives/systemd-devel/2014-August/022537.html There's been similar issues with their dhcp. pgpRjcjiVttxV.pgp Description: PGP signature
Re: Linux: concerns over systemd [OT]
Once upon a time, valdis.kletni...@vt.edu valdis.kletni...@vt.edu said: Actually, systemd 216 will cram systemd-timesyncd down your throat even if you had ntpd installed. Yeah, I think a lot of the upset with systemd is not so much with the core daemon that runs as PID 1, but with the massive scope creep that the systemd project appears to have. Why should a project centered around an init system start reimplementing system logging, network management (and a DHCP client), clock management, etc.? Lennart's attitude was pretty much why would anybody want to run ntpd when they have our SNTP implementation: Wow, maybe because SNTP is inferior to an actual NTP daemon in just about every way? -- Chris Adams c...@cmadams.net
Re: Linux: concerns over systemd [OT]
On Wed, Oct 22, 2014 at 3:34 PM, Randy Bush ra...@psg.com wrote: the vast majority of negative tongue wagging regarding systemd is ill informed. can we skip the ad homina and leave that for the systemd dev fora? I don't think that it's an ad homina attack, as it's pretty clear that many of the people commenting have not spent a significant time using systemd so many of their comments are based on what they've read on the Internet, not from practical experience with systemd. does systemd have growing pains? definitely. are some egos involved? sure. can systemd be far reaching? yes, is such reach mandated? no. use the units you want and disregard the rest. how does this work out in practice? at install, can i choose whether systemd is used for X as opposed for the separate component? can i template such choices for cluster deployment with the usual tools? I think that Debian's plan to allow multiple init systems (irregardless of which one is default) is a bad plan. The non-default ones won't get any love - at some point they'll just stop working (or indeed, work at all). Allowing choice of components is a good thing at one level (e.g. sendmail vs. postfix vs. exim). I really don't care (and don't really even remeber) which SMTP server is installed by default on my systems because my configuration management system makes sure that the SMTP server that I prefer is installed and configured the way I want it once the system is up and running. For something like PID 1, each distribution should make a choice and stick with it. I really couldn't care what Debian's init system is, as I don't use Debian (never have, at least not when I have had a choice). If Debian goes through with the switch to systemd, they won't gain me as a user as there are a host of other reasons that I prefer something other than Debian (or Debian-derived) distributions. If a group of people fork Debian because of systemd, more power to them. -- Jeff Ollie
Re: Linux: concerns over systemd [OT]
On Wed, Oct 22, 2014 at 3:48 PM, valdis.kletni...@vt.edu wrote: On Wed, 22 Oct 2014 19:35:51 -, David Ford said: into a common bus. not everything in systemd is a requirement to run it. just because a unit is offered for dhcp or ntp, doesn't mean you are required to use it. Actually, systemd 216 will cram systemd-timesyncd down your throat even if you had ntpd installed. Oh really? From my Fedora 21 (alpha) system at work: # rpm -q systemd systemd-216-5.fc21.x86_64 # systemctl status systemd-timesyncd.service ● systemd-timesyncd.service - Network Time Synchronization Loaded: loaded (/usr/lib/systemd/system/systemd-timesyncd.service; disabled) Active: inactive (dead) Docs: man:systemd-timesyncd.service(8) # systemctl status ntpd.service ● ntpd.service - Network Time Service Loaded: loaded (/usr/lib/systemd/system/ntpd.service; enabled) Active: active (running) since Thu 2014-10-16 16:06:22 CDT; 6 days ago Main PID: 1438 (ntpd) CGroup: /system.slice/ntpd.service └─1438 /usr/sbin/ntpd -u ntp:ntp -g I'm not sure what NTP service is installed by default in Fedora 21+ (which will ship with systemd 216), as this system has been upgraded from previous Fedora versions, but as you can see it's perfectly possible to run ntpd on a system that used systemd 216. https://bugzilla.redhat.com/show_bug.cgi?id=1136905 https://mail.gnome.org/archives/distributor-list/2014-September/msg2.html Lennart's attitude was pretty much why would anybody want to run ntpd when they have our SNTP implementation: A vast oversimplification of Lennart's point. Basically you left off if you really want to run chronyd or ntpd service go right ahead, we're just not going to have code in systemd to do that for you. http://lists.freedesktop.org/archives/systemd-devel/2014-August/022537.html There's been similar issues with their dhcp. The bug here is really timedatectl (a component of systemd) isn't going to manage chronyd or ntpd (third party packages), and that made a Gnome control panel (which used timedatectl under the hood) report that network time synchronization wasn't enabled. If you don't want timedated then it's perfectly possible to disable timedated and use something else. If someone cares enough, I'm sure that the Gnome control panel will get updated so that I can manage chronyd or ntpd itself. -- Jeff Ollie
Re: Linux: concerns over systemd [OT]
On Wed, Oct 22, 2014 at 9:17 PM, Jeffrey Ollie j...@ocjtech.us wrote: I think that Debian's plan to allow multiple init systems (irregardless of which one is default) is a bad plan. The non-default ones won't get any love - at some point they'll just stop working (or indeed, work at all). Indeed. I believe that point was made during the debian technical committee discussions by one of the members of the TC (Russ, I think, although it was such a long discussion it could have been one of the other participants).
Re: Linux: concerns over systemd [OT]
Bah, boot speed; On my server, boot is slow down by hardware initialization. The soft side is quite low. But the point is not makes things faster from 15 to 14 sec is useless. The point is : it's good, but at what price ? As you said, there were many improvements over the past. What was the clean bit cost ? None but benefits, right ? What about fs logs ? Does it have a cost ? If systemd is just about time, it will be fine. But why trying to recreate (ans thus, squeeze) some old daemons like cron or syslog ? Both of them are doing a perfect job. Can I use systemd without any of journald stuff ? If not, then the 1sec speedup is far too expensive. On 22/10/2014 20:31, Barry Shein wrote: On October 21, 2014 at 16:43 morrowc.li...@gmail.com (Christopher Morrow) wrote: On Tue, Oct 21, 2014 at 4:10 PM, Andrew Sullivan asulli...@dyn.com wrote: On Tue, Oct 21, 2014 at 03:11:55PM -0400, Barry Shein wrote: But for example some of my servers boot in seconds. One is reminded of a mail, included in the Preface to _The UNIX-HATERS Handbook_, available at it's really not clear to me that 'reboots in seconds' is a thing to optimize... The unix community has exerted great amounts of effort over the decades to speed up reboot, particularly after crashes but also planned. Perhaps you don't remember the days when an fsck was basically mandatory and could take 15-20 minutes on a large disk. Then we added the clean bit (disk unmounted cleanly, no need for fsck), reorg'd the file system layout to speed up fsck considerably and make it more reliable/recoverable, added journaled file systems which really sped things up often eliminating the need to fsck after a crash entirely and recovering in seconds, various attempts to figure out the dependency graph of servers and services which need to be started so they could be started in parallel where dependencies are met, etc. And learned how to do hot failover and master/slave servers etc. And you whisk all that away with it's not really clear to me that 'reboots in seconds' is a think to be optimized To me that's like saying it's not important to try to design so one can recover from a network outage in seconds. Anyhow, if it's not clear: I disagree. I suppose the win is: Is the startup/shutdown process clear, conscise and understandable at 3am local time? followed by: Can I adjust my startup processes to meet my needs easily and without finding a phd in unix? If systemd is simply a change in how I think about /etc/init.d/* and /etc/rc?.d/* cool, if it's more complexity and less EASY flexibility then it's a fail. Actually, much of that is less important except perhaps to a hobbyist. You only have to get the startup/shutdown process etc right once in a while and generally during a planned outage. Recovering from a failure or going back into service quickly after a planned outage is critical and can be critical at any time. Obviously one can appeal to extremum but what you say doesn't make sense to me. At any rate, you are disputing a huge, decades long, and widely fought battle. It's certainly not my opinion.
Re: Linux: concerns over systemd [OT]
On Wed, Oct 22, 2014 at 4:48 PM, na...@jack.fr.eu.org wrote: Bah, boot speed; On my server, boot is slow down by hardware initialization. The soft side is quite low. But the point is not makes things faster from 15 to 14 sec is useless. The point is : it's good, but at what price ? I agree that boot speed is a red herring. Booting in a highly dynamic environment is really one of systemd's key achievements. True, this is most apparent in systems like laptops that can be docked/undocked, tend to have a lot of USB devices added/removed, etc. But servers have these same issues, if only in lesser degree. sysvinit generally had to wait for a fixed interval for all hardware to finish initializing. It was a little more sophisticated than one init script having a sleep X command in it, but not much. systemd is able to start services as soon as all of the hardware it needs is ready, rather than waiting arbitrary amounts of time and then possibly failing anyway because it didn't wait long enough. One main example of this is a large RAID array or LVM volume that's used for data storage (not the OS). systemd can let parts of the system boot that don't require the data storage to be present start up while waiting for all of the data storage drives to spin up and get assembled. As you said, there were many improvements over the past. What was the clean bit cost ? None but benefits, right ? What about fs logs ? Does it have a cost ? If systemd is just about time, it will be fine. But why trying to recreate (ans thus, squeeze) some old daemons like cron or syslog ? Both of them are doing a perfect job. Syslog didn't capture the stdout/stderr from daemon start up. Syslog did only a mediocre job of capturing dmesg from early boot. If you have access to a system running systemd/journald run journalctl -k -b. That'll show you all of the kernel messages since the last boot. At least on the RHEL systems that I have access to, /var/log/dmesg doesn't timestamp the lines in the file, making it difficult to impossible to correlate with other log lines. The kernel messages in /var/log/messages are timestamped with the time that rsyslog was able to (finally) start and pull the messages out of the kernel message buffer. Can I use systemd without any of journald stuff ? I wouldn't want to. If not, then the 1sec speedup is far too expensive. The boot speed up is a nice benefit, but not the only (or even the best) reason to use systemd. -- Jeff Ollie
Self destruction in open source systems (was Re: Linux: concerns over systemd [OT])
On 10/22/2014 06:01, Randy Bush wrote: Which leads me to ask - those of you running server farms - what distros are popular these days, for server-side operations? been running bsd forever. but moving to debian and ganeti, as bsd does not host virtualization. Simply not true; http://bhyve.org/ It is a bit immature compared to Xen+Ganeti or something like that. apologies. i thought we were talking about production systems. my mistake. This may take us far enough away from BGP and prefix lengths to draw moderator fire, but it is a topic I would like to pursue somewhere. What is it about seems to mandate that they develop a fatal rot from the inside out? I don't have anything to add to the xX OSes (they Were Not Allowed when I was an active admin), but I have of late been grappling with what to do about replacing Firefox and Thunderbird in the four machines I have left in my world. Early on I developed a fondness for Mosaic as a simple get-the-job done tool to add to elm and news to get the work done. Netscape did a nice integration and I went with that until it went away. Now I have Thunderbird and Firefox--from people who are committed to the notion that if it works, it must be replaced. If people like it, it must be redesigned. If it is stable, it must be updated. If there is a useless, senseless feature somewhere in the world, these products must be revised to make that feature the focus. -- The unique Characteristics of System Administrators: The fact that they are infallible; and, The fact that they learn from their mistakes. Quis custodiet ipsos custodes
Re: Self destruction in open source systems (was Re: Linux: concerns over systemd [OT])
On Wed, Oct 22, 2014 at 4:58 PM, Larry Sheldon larryshel...@cox.net wrote: Now I have Thunderbird and Firefox--from people who are committed to the notion that if it works, it must be replaced. If people like it, it must be redesigned. If it is stable, it must be updated. If there is a useless, senseless feature somewhere in the world, these products must be revised to make that feature the focus. And where is my new 1967 VW Microbus? That's all you need if you compile it with --add-heater-fan. So I had to upgrade to a 1998 Volvo V70 wagon. Don't know where I'm going to get a new one when this one wears out. Damn kids, GET OFF MY LAWN! I actually feel with your there, Larry. I really like the *nixes because of the great app store with things like ls, grep, sed, cc and ssh. It's also why for most things I still use one of the BSDs. (Should we call /usr/ports an app store now?) -- Joe Hamelin, W7COM, Tulalip, WA, 360-474-7474
Re: Self destruction in open source systems (was Re: Linux: concerns over systemd [OT])
Joe Hamelin wrote: On Wed, Oct 22, 2014 at 4:58 PM, Larry Sheldon larryshel...@cox.net wrote: Now I have Thunderbird and Firefox--from people who are committed to the notion that if it works, it must be replaced. If people like it, it must be redesigned. If it is stable, it must be updated. If there is a useless, senseless feature somewhere in the world, these products must be revised to make that feature the focus. And where is my new 1967 VW Microbus? That's all you need if you compile it with --add-heater-fan. So I had to upgrade to a 1998 Volvo V70 wagon. Don't know where I'm going to get a new one when this one wears out. Well hey, nobody's made a good 4WD station wagon since Toyota stopped making them. At one point it seemed like every 3rd person in the BBN parking lot had one, and we all drove them into the ground for lack of a replacement. Sigh... -- In theory, there is no difference between theory and practice. In practice, there is. Yogi Berra
Re: Linux: concerns over systemd [OT]
On Wed, Oct 22, 2014 at 1:31 PM, Barry Shein b...@world.std.com wrote: [snip] The unix community has exerted great amounts of effort over the decades to speed up reboot, particularly after crashes but also planned. Perhaps you don't remember the days when an fsck was basically mandatory and could take 15-20 minutes on a large disk. Then we added the clean bit (disk unmounted cleanly, no need for [snip] And you whisk all that away with it's not really clear to me that 'reboots in seconds' is a think to be optimized False dilemma. Optimizing reboot time down from 20 minutes to 1 minute is a significantly meaningful improvement; it's literally a 85% reduction in time spent during each boot process from the original time. Reducing boot time from 20 minutes to 10 seconds is not significantly better than reducing it to 1 minute. A different choice of tradeoffs is more appropriate to different kinds of systems, depending on their use case (Desktop vs Server)! Especially, when the method of reduction is subject to diminishing returns and increasing fragility or increasing complexity -- greater risk that something is breaking or more potential for unreliability is introduced into the startup process. Also, you may very well spend more time booting your system in order to troubleshoot, the fact that some applications are starting up in an unexpected order resulting in some issue. To me that's like saying it's not important to try to design so one can recover from a network outage in seconds. If you need to ensure that a service is not disrupted for more than seconds, then reboot is not the answer. It is some form of clustering. Reboot as a troubleshooting procedure is for desktops. 10 seconds from power on to user interface for desktops, will meaningfully improve the user experience, but not for servers. For servers, you ideally want to take the misbehaving node out of service and let its failover partner takeover. -- -JH
Re: Linux: concerns over systemd [OT]
On Wed, Oct 22, 2014 at 9:48 PM, Jimmy Hess mysi...@gmail.com wrote: On Wed, Oct 22, 2014 at 1:31 PM, Barry Shein b...@world.std.com wrote: And you whisk all that away with it's not really clear to me that 'reboots in seconds' is a think to be optimized False dilemma. [ snip ] 10 seconds from power on to user interface for desktops, will meaningfully improve the user experience, but not for servers. It's a false dilemma only if you're thinking about traditional physical servers. Consider: 1) What if you're spinning up several thousand Hadoop nodes on AWS or GCE so that you can do some sort of big data operation. 2) What if PewDiePie just mentioned one of your products in a video and you need to quickly scale up the number of backend servers to handle the load. I'm sure that there are many other scenarios that I could devise where a fast server boot time was important. -- Jeff Ollie
Re: Linux: concerns over systemd [OT]
On Wed, Oct 22, 2014 at 09:48:51PM -0500, Jimmy Hess wrote: Optimizing reboot time down from 20 minutes to 1 minute is a significantly meaningful improvement; it's literally a 85% reduction in time spent during each boot process from the original time. if reducing boot time from 20 minutes down to 1 minute, in a server environment, is a serious issue for you, maybe you should be looking at why you need to reboot so often? i'm somewhat puzzled by the fanboi mantra of i've been running whizzy weasel and have 1574 days of uptime, which has now been supplanted by geez, i have to wait 3 minutes every time i reboot this thing. running ntp, dhcp, dns, smtp, imap, http, that's what we do in serverland. and in addition to that, we need to run whatever the latest and greatest piece of crap that's being touted on slashdot (redis, mongo, couchbase, elasticsearch, anything that uses ruby/forever). we generally don't have alot of say in what we have to run because the fanboi's run the media, and management tends to give media more credence than the decades of experience they have in-house. that's how linux made it into the server environment in the first place. systemd sounds like a really useful thing if you are running a desktop system. as far as booting up a server, to run services, and to keep those services running, the init.d/rc.d/etc systems do a good job, and its generally not that hard to add/modify if you are half-assed competent. before criticizing people for being afraid of new technology, make sure that you yourself are not afraid of existing technology. --jim -- Jim Mercer Reptilian Research j...@reptiles.org+1 416 410-5633 He who dies with the most toys is nonetheless dead
Re: Linux: concerns over systemd [OT]
On 10/22/2014 23:02, Jim Mercer wrote: if reducing boot time from 20 minutes down to 1 minute, in a server environment, is a serious issue for you, maybe you should be looking at why you need to reboot so often? That is the question I have been asking myself. Back in the day we took it a a failure if a reboot happened. (I remember discussions about needing to reboot to keep counters from overflowing. I thought programming for counter wrap was a better idea.) -- The unique Characteristics of System Administrators: The fact that they are infallible; and, The fact that they learn from their mistakes. Quis custodiet ipsos custodes
Re: Linux: concerns over systemd [OT]
On Wed, 22 Oct 2014, Larry Sheldon wrote: That is the question I have been asking myself. Back in the day we took it a a failure if a reboot happened. (I remember discussions about needing to reboot to keep counters from overflowing. I thought programming for counter wrap was a better idea.) When I see a machine with a long uptime I ask myself: 1. When was the last kernel update? 2. Do we know it would come back cleanly when it is rebooted? 3. What would happen to the site if that machine went offline? -- Simon Lyall | Very Busy | Web: http://www.simonlyall.com/ To stay awake all night adds a day to your life - Stilgar
Re: Linux: concerns over systemd [OT]
Ok. As a highly on- list-topic example of why distrust is called for... Without referring to the systemd source code*, does anyone know what systemd uses to select between networking subsystems (i.e. NetworkManager, the new standard as of RHEL 7, vs /etc/ sysconfig/network-scripts/, etc.). NetworkManager is default but disableable and it magically falls back to network-scripts dir, but the fallback is nearly undocumented and the selection behavior appears completely undocumented. If by some chance you do know this, where did you come by that knowledge? Hopefully with URLs. (* don't bother telling me to read the source. I'm reading...) If I cannot find credible documentation of this, as networking person as well as enterprise sysadmin, this is a Problem.) George William Herbert Sent from my iPhone
Re: Linux: concerns over systemd [OT]
On 10/22/14 9:29 PM, Larry Sheldon wrote: On 10/22/2014 23:02, Jim Mercer wrote: if reducing boot time from 20 minutes down to 1 minute, in a server environment, is a serious issue for you, maybe you should be looking at why you need to reboot so often? That is the question I have been asking myself. Back in the day we took it a a failure if a reboot happened. (I remember discussions about needing to reboot to keep counters from overflowing. I thought programming for counter wrap was a better idea.) Back over here in router-land when the cat65k vss pair went down it's at least 20 minutes before it's back. If you enjoy what may be the longest minutes of your life try tripping over a bug that takes out two pairs and a whole pop. Having something come back quickly is part of having flexibility since things do go pear-shaped, and responding to outages rather than not having them is not tick box we get to decline. Today my Arista reloada in about 220 seconds which is tolerable but if it were half that I'd be even happier. signature.asc Description: OpenPGP digital signature
Re: Linux: concerns over systemd [OT]
I've done a fair amount of hand-to-hand combat with systemd. When it's good it's good, tho not always apparent why it's good. But for example some of my servers boot in seconds. When it's bad it can be painful and incredibly opaque and a huge time sink. Googling for suggestions I've found several threads where the co-author (Poettering) jumps in usually to be annoyingly arrogant (I'm sure he's very bright and good to children and pets and overworked) responding with comments like why don't you just read your logs and not bother this list or similar (that was paraphrased.) The logs are, in my experience, almost always useless or nearly so, mumble failed to start basically. I'm not the only one: http://www.muktware.com/2014/04/linus-torvalds-happy-systemd-author-kay-sievers/25151 It also resists tools like strace because it tends to do things by IPC. In one extreme case I just reworked an /etc/init.d script to avoid systemd (not use the various /etc/rc.foo files), mostly just hit it with a sledgehammer and put fixing that on my TODO list. Unfortunately I am mortal and have limited time on this earth. My experience as I said is mixed, hard cases are very hard where they really seem like they shouldn't be (just tell me roughly what you're trying to do rather than just fail, eg, via some debug enable), most are just your usual oops it wants this or that situations. I don't think I'd want to revert to sysvinit, systemd seems architecturally superior. But it needs a lot more transparency and some attempt to gather common problems -- like why is it hanging asking for a password on the console when I can't see why it thinks it needs one? -- and FAQ them with real answers or add some code/configuration to fix that (never ask for a password in this script OK? And no --no-ask-password isn't fixing this so stop repeating that answer!) -- -Barry Shein The World | b...@theworld.com | http://www.TheWorld.com Purveyors to the Trade | Voice: 800-THE-WRLD| Dial-Up: US, PR, Canada Software Tool Die| Public Access Internet | SINCE 1989 *oo*
Re: Linux: concerns over systemd [OT]
On Tue, Oct 21, 2014 at 03:11:55PM -0400, Barry Shein wrote: But for example some of my servers boot in seconds. One is reminded of a mail, included in the Preface to _The UNIX-HATERS Handbook_, available at http://www.art.net/~hopkins/Don/unix-haters/preface.html. Apparently, things really are going to get a lot worse before they get worse. Best regards, A -- Andrew Sullivan Dyn, Inc. asulli...@dyn.com v: +1 603 663 0448
Re: Linux: concerns over systemd [OT]
On Tue, Oct 21, 2014 at 4:10 PM, Andrew Sullivan asulli...@dyn.com wrote: On Tue, Oct 21, 2014 at 03:11:55PM -0400, Barry Shein wrote: But for example some of my servers boot in seconds. One is reminded of a mail, included in the Preface to _The UNIX-HATERS Handbook_, available at it's really not clear to me that 'reboots in seconds' is a thing to optimize... I suppose the win is: Is the startup/shutdown process clear, conscise and understandable at 3am local time? followed by: Can I adjust my startup processes to meet my needs easily and without finding a phd in unix? If systemd is simply a change in how I think about /etc/init.d/* and /etc/rc?.d/* cool, if it's more complexity and less EASY flexibility then it's a fail. -chris
Re: Linux: concerns over systemd [OT]
Christopher Morrow wrote: On Tue, Oct 21, 2014 at 4:10 PM, Andrew Sullivan asulli...@dyn.com wrote: On Tue, Oct 21, 2014 at 03:11:55PM -0400, Barry Shein wrote: But for example some of my servers boot in seconds. One is reminded of a mail, included in the Preface to _The UNIX-HATERS Handbook_, available at it's really not clear to me that 'reboots in seconds' is a thing to optimize... I suppose the win is: Is the startup/shutdown process clear, conscise and understandable at 3am local time? followed by: Can I adjust my startup processes to meet my needs easily and without finding a phd in unix? If systemd is simply a change in how I think about /etc/init.d/* and /etc/rc?.d/* cool, if it's more complexity and less EASY flexibility then it's a fail. You guys REALLY don't want to wade into the swamp on debian-users -- the place is full of systemd fanboys and apologists, and anybody who raises real operational concerns resulting from the switch in default init systems. I'm really pining for a LISP Machine right about now. Miles Fidelman -- In theory, there is no difference between theory and practice. In practice, there is. Yogi Berra
Re: Linux: concerns over systemd [OT]
The thing that I don't understand about systemd is how it managed to get *EVERY SINGLE DISTRIBUTION'S RELEASE MANAGER* on board in less than a year, given how thoroughly it violates the Unix philosophy, and how poorly documented it is -- to the point where you can't even run sysvinit anymore unless you're willing to build initscripts by hand, since packages don't even include them anymore. Does Poettering have compromising photographs of all these guys in a puppy pile at a Linuxcon somewhere? Cheers, -- jra - Original Message - From: Barry Shein b...@world.std.com To: Israel G. Lugo israel.l...@lugosys.com Cc: nanog@nanog.org Sent: Tuesday, October 21, 2014 3:11:55 PM Subject: Re: Linux: concerns over systemd [OT] I've done a fair amount of hand-to-hand combat with systemd. When it's good it's good, tho not always apparent why it's good. But for example some of my servers boot in seconds. When it's bad it can be painful and incredibly opaque and a huge time sink. Googling for suggestions I've found several threads where the co-author (Poettering) jumps in usually to be annoyingly arrogant (I'm sure he's very bright and good to children and pets and overworked) responding with comments like why don't you just read your logs and not bother this list or similar (that was paraphrased.) The logs are, in my experience, almost always useless or nearly so, mumble failed to start basically. I'm not the only one: http://www.muktware.com/2014/04/linus-torvalds-happy-systemd-author-kay-sievers/25151 It also resists tools like strace because it tends to do things by IPC. In one extreme case I just reworked an /etc/init.d script to avoid systemd (not use the various /etc/rc.foo files), mostly just hit it with a sledgehammer and put fixing that on my TODO list. Unfortunately I am mortal and have limited time on this earth. My experience as I said is mixed, hard cases are very hard where they really seem like they shouldn't be (just tell me roughly what you're trying to do rather than just fail, eg, via some debug enable), most are just your usual oops it wants this or that situations. I don't think I'd want to revert to sysvinit, systemd seems architecturally superior. But it needs a lot more transparency and some attempt to gather common problems -- like why is it hanging asking for a password on the console when I can't see why it thinks it needs one? -- and FAQ them with real answers or add some code/configuration to fix that (never ask for a password in this script OK? And no --no-ask-password isn't fixing this so stop repeating that answer!) -- -Barry Shein The World | b...@theworld.com | http://www.TheWorld.com Purveyors to the Trade | Voice: 800-THE-WRLD | Dial-Up: US, PR, Canada Software Tool Die | Public Access Internet | SINCE 1989 *oo* -- Jay R. Ashworth Baylink j...@baylink.com Designer The Things I Think RFC 2100 Ashworth Associates http://www.bcp38.info 2000 Land Rover DII St Petersburg FL USA BCP38: Ask For It By Name! +1 727 647 1274
Re: Linux: concerns over systemd [OT]
- Original Message - From: Capi c...@lugosys.com On 10/21/2014 11:29 PM, Jay Ashworth wrote: The thing that I don't understand about systemd is how it managed to get *EVERY SINGLE DISTRIBUTION'S RELEASE MANAGER* on board in less than a year, given how thoroughly it violates the Unix philosophy, and how poorly documented it is Not *every single* distribution... I had meant to put an asterisk on that. I'm glad to be using Gentoo Linux at home for the last 10 years... They've adopted OpenRC, which is much less invasive, works with an existing init (possibly sysv) and uses the friendly shell scripts we're all used to. Ok, but how does it handle providing initscripts? I gather any upstreams which used to provide them aren't anymore... Cheers, -- jra -- Jay R. Ashworth Baylink j...@baylink.com Designer The Things I Think RFC 2100 Ashworth Associates http://www.bcp38.info 2000 Land Rover DII St Petersburg FL USA BCP38: Ask For It By Name! +1 727 647 1274
Re: Linux: concerns over systemd [OT]
On 21/10/14 23:55, Jay Ashworth wrote: Ok, but how does it handle providing initscripts? I gather any upstreams which used to provide them aren't anymore... It's Gentoo: You should write your own is the most likely answer. -- Tom
Re: Linux: concerns over systemd [OT]
Probably a lot of it has to do with: - we're merging udev and a bunch of other things into systemd - you want GNOME to work, you'd better use systemd - Canonical (Ubuntu) DIDN'T commit to udev until Debian made the decision - they would have kept going with upstart, but when Debian committed, they decided they didn't want to support a now-orphaned init system - Gentoo supports systemd as an option, it's fork funtoo doesn't - Slackware doesn't Miles Fidelman Jay Ashworth wrote: The thing that I don't understand about systemd is how it managed to get *EVERY SINGLE DISTRIBUTION'S RELEASE MANAGER* on board in less than a year, given how thoroughly it violates the Unix philosophy, and how poorly documented it is -- to the point where you can't even run sysvinit anymore unless you're willing to build initscripts by hand, since packages don't even include them anymore. Does Poettering have compromising photographs of all these guys in a puppy pile at a Linuxcon somewhere? Cheers, -- jra - Original Message - From: Barry Shein b...@world.std.com To: Israel G. Lugo israel.l...@lugosys.com Cc: nanog@nanog.org Sent: Tuesday, October 21, 2014 3:11:55 PM Subject: Re: Linux: concerns over systemd [OT] I've done a fair amount of hand-to-hand combat with systemd. When it's good it's good, tho not always apparent why it's good. But for example some of my servers boot in seconds. When it's bad it can be painful and incredibly opaque and a huge time sink. Googling for suggestions I've found several threads where the co-author (Poettering) jumps in usually to be annoyingly arrogant (I'm sure he's very bright and good to children and pets and overworked) responding with comments like why don't you just read your logs and not bother this list or similar (that was paraphrased.) The logs are, in my experience, almost always useless or nearly so, mumble failed to start basically. I'm not the only one: http://www.muktware.com/2014/04/linus-torvalds-happy-systemd-author-kay-sievers/25151 It also resists tools like strace because it tends to do things by IPC. In one extreme case I just reworked an /etc/init.d script to avoid systemd (not use the various /etc/rc.foo files), mostly just hit it with a sledgehammer and put fixing that on my TODO list. Unfortunately I am mortal and have limited time on this earth. My experience as I said is mixed, hard cases are very hard where they really seem like they shouldn't be (just tell me roughly what you're trying to do rather than just fail, eg, via some debug enable), most are just your usual oops it wants this or that situations. I don't think I'd want to revert to sysvinit, systemd seems architecturally superior. But it needs a lot more transparency and some attempt to gather common problems -- like why is it hanging asking for a password on the console when I can't see why it thinks it needs one? -- and FAQ them with real answers or add some code/configuration to fix that (never ask for a password in this script OK? And no --no-ask-password isn't fixing this so stop repeating that answer!) -- -Barry Shein The World | b...@theworld.com | http://www.TheWorld.com Purveyors to the Trade | Voice: 800-THE-WRLD | Dial-Up: US, PR, Canada Software Tool Die | Public Access Internet | SINCE 1989 *oo* -- In theory, there is no difference between theory and practice. In practice, there is. Yogi Berra
Re: Linux: concerns over systemd [OT]
On 10/21/2014 11:55 PM, Jay Ashworth wrote: - Original Message - From: Capi c...@lugosys.com Whoops, used the wrong alias to reply. Not *every single* distribution... I had meant to put an asterisk on that. My remark was meant to be tongue-in-cheek :) Ok, but how does it handle providing initscripts? I gather any upstreams which used to provide them aren't anymore... The Gentoo devs take care of that. I presume they reuse what they can from upstream... They do a lot of hard work (sometimes more work than they have the manpower for, unfortunately). I remember, for example, back in KDE 3.5 days they were already dividing the upstream KDE mega packages (kde-games, kde-office) into individual packages, so you could choose specific programs instead of 300 MB bundles.
Re: Linux: concerns over systemd [OT]
On Tue, 21 Oct 2014 18:29:44 -0400, Jay Ashworth j...@baylink.com wrote: The thing that I don't understand about systemd is how it managed to get *EVERY SINGLE DISTRIBUTION'S RELEASE MANAGER* on board... It's spelled Red Hat. Add in GNOME and debian (et. al.) is backed into a corner. Red Hat is soo f'ing big, pretty much every project under the sun is going to stop maintaining scripts in favor of systemd.
Re: Linux: concerns over systemd [OT]
On 10/21/2014 11:59 PM, Tom Hill wrote: On 21/10/14 23:55, Jay Ashworth wrote: Ok, but how does it handle providing initscripts? I gather any upstreams which used to provide them aren't anymore... It's Gentoo: You should write your own is the most likely answer. Actually, not at all; although I realize that's a very common misconception. Gentoo Linux is, unfortunately, often associated with the whole gcc -O9000 -msuperfast -fwtf wow-look-at-me crowd. It's true that some people who use Gentoo go on and rave about how many nanoseconds they were able to shave off of their boot time, or how many obscure undocumented GCC options they managed to squeeze in without a compile error. I suppose the flexible nature of Gentoo is appealing to those who like to look cool and show off how they can watch the compiler do its thing. However, that's not at all what the distribution is about. Gentoo is about flexibility and choice. It's got a steepish learning curve, yes, but the documentation is very good; sadly, much of it was lost a few years ago, due to a bad mishap on the community Gentoo Wiki server, apparently without any backups. Back in the day, if I wanted to learn about Samba, I'd Google howto linux samba and Gentoo's Wiki would usually be among the first 3 hits. Their devs take stability very seriously; it's a rolling distro, but there is still a reasonable stabilization period for each package as new versions come out, during which any open bugs may hold up the package until they're fixed. It's all about choice. In my view, Gentoo is no better or worse than Debian, Red Hat, or Ubuntu. Different species, they all make for a better ecosystem.
Re: Linux: concerns over systemd [OT]
Israel G. Lugo wrote: On 10/21/2014 11:59 PM, Tom Hill wrote: On 21/10/14 23:55, Jay Ashworth wrote: Ok, but how does it handle providing initscripts? I gather any upstreams which used to provide them aren't anymore... It's Gentoo: You should write your own is the most likely answer. Actually, not at all; although I realize that's a very common misconception. Gentoo Linux is, unfortunately, often associated with the whole gcc -O9000 -msuperfast -fwtf wow-look-at-me crowd. It's true that some people who use Gentoo go on and rave about how many nanoseconds they were able to shave off of their boot time, or how many obscure undocumented GCC options they managed to squeeze in without a compile error. I suppose the flexible nature of Gentoo is appealing to those who like to look cool and show off how they can watch the compiler do its thing. However, that's not at all what the distribution is about. Gentoo is about flexibility and choice. It's got a steepish learning curve, yes, but the documentation is very good; sadly, much of it was lost a few years ago, due to a bad mishap on the community Gentoo Wiki server, apparently without any backups. Back in the day, if I wanted to learn about Samba, I'd Google howto linux samba and Gentoo's Wiki would usually be among the first 3 hits. Their devs take stability very seriously; it's a rolling distro, but there is still a reasonable stabilization period for each package as new versions come out, during which any open bugs may hold up the package until they're fixed. It's all about choice. In my view, Gentoo is no better or worse than Debian, Red Hat, or Ubuntu. Different species, they all make for a better ecosystem. Given the state of things, though, I'm more-and-more considering Linux from Scratch. I find that I install enough from upstream source that packaging systems (and out-of-date packages) are less and less useful. Probably easier to set up Chef or Puppet and Jenkins to just keep the overall system current - and the heck with all this distro nonsense. Cheers, Miles Fidelman -- In theory, there is no difference between theory and practice. In practice, there is. Yogi Berra
Re: Linux: concerns over systemd [OT]
On 22/10/14 00:57, Israel G. Lugo wrote: Gentoo is about flexibility and choice. It's got a steepish learning curve, yes, but the documentation is very good; sadly, much of it was lost a few years ago, due to a bad mishap on the community Gentoo Wiki server, apparently without any backups. Back in the day, if I wanted to learn about Samba, I'd Google howto linux samba and Gentoo's Wiki would usually be among the first 3 hits. Their devs take stability very seriously; it's a rolling distro, but there is still a reasonable stabilization period for each package as new versions come out, during which any open bugs may hold up the package until they're fixed. I certainly remember this, and miss it. The Gentoo documentation, and indeed the experience of compiling everything, was excellent. I still miss some of the tools that Gentoo had in Debian/CentOS (and the stage3 live CD is still my goto 'system rescue tool' :)) But.. I don't use it any more for anything serious. It's too much upkeep, and when the the included/maintained rc scripts for some package do inevitably fail to catch a corner case -- far more likely if you're using an overlay -- then you're left with little choice but to start modifying/writing your own. It's all about choice. In my view, Gentoo is no better or worse than Debian, Red Hat, or Ubuntu. Different species, they all make for a better ecosystem. I was mildly unfair in the way my response was worded, but I do hold that the Gentoo way of doing things is much simpler than that of other distributions. This was, in my experience, a double-edged sword. YMMV, etc. -- Tom
Re: Linux: concerns over systemd [OT]
- Original Message - From: Ricky Beam jfb...@gmail.com On Tue, 21 Oct 2014 18:29:44 -0400, Jay Ashworth j...@baylink.com wrote: The thing that I don't understand about systemd is how it managed to get *EVERY SINGLE DISTRIBUTION'S RELEASE MANAGER* on board... It's spelled Red Hat. Add in GNOME and debian (et. al.) is backed into a corner. Red Hat is soo f'ing big, pretty much every project under the sun is going to stop maintaining scripts in favor of systemd. GNOME is probably the linchpin. But it's not just RH. It's Debian, and by extension *buntu, and SuSE, and at least one other major independent parent distro that I can't think of just now... And as far as I know, it's done; SuSE packages already largely don't even include initscripts. Cheers, -- jra -- Jay R. Ashworth Baylink j...@baylink.com Designer The Things I Think RFC 2100 Ashworth Associates http://www.bcp38.info 2000 Land Rover DII St Petersburg FL USA BCP38: Ask For It By Name! +1 727 647 1274