Fedora :: Analyze The Startup Time Of Your OS With Systemd?
Aug 17, 2011
We have seen with the release of Fedora 15, systemd is the new start-up services, which will replace SysVinit and Upstart on most Linux distributions. Here are some tips based on the order systemd-analyze, for analyzing the startup time of your OS.On commence avec l'option time, qui affiche le temps total de d�marrage de votre syst�me : (One starts with l' option time, which posts the total time of starting of your system: )
$ systemd-analyze time
Startup finished in 6385ms (kernel) + 3228ms (initrd) + 49335ms (userspace) = 58949ms
I got a message today telling me that there where 78 updates available. One of them was Kernel 126.96.36.199-213.fc11.i586 and when I installed all the updates and rebooted, my startup time was very slow. And when I checked the boot.log I noticed this:
Because of my problem here:[URL]..I'd like to replace upstart with systemd on F14. Is this as simple as a yum install? In the wiki is it's referred to as a "technology preview" for F14, which is meaningless to me; there are no further notes about installing or enabling it.
What's the best way to get a list of the start order of services from systemd? Yes, I've tried
This looks like the output I would expect, except that if that list is the system service start order, then my system is hosed. I certainly hope that my system is not starting nfs service before all the /usr/export filesystems are mounted.
systemd --test --system
but that generates over 22 thousand lines of gibberish! I filtered that with
systemd --test --system | grep -- -> | grep Unit | grep -i service
And it's less gibberish, but it's ALMOST the reverse order of the first command, and still pretty crazy.Ultimately, I need to discover, and fix the order of the services started by systemd, because they were mangled during the upgrade from Fedora 14 I hope that "systemctl --order" is either not the real deal, or I'm reading it wrong. Because if it is, I'm afraid I have a huge problem to fix.
I've proven that the the D in systemD is for DISASTER.At first, I thought that the dozens of problems I had with startup, shutdown, and services, was due to my upgrading from Fedora 14. But now that I've tried a fresh install, I have a new set of different problems. Basically DBUS broke after I installed Samba, and no combo of reinstalling either of them fixed anything. So here goes my third Fedora 15 install.
During a fresh install, can I de-select systemd and select upstart and sysv instead? systemd is not visible in the GUI installer. Do I need some Kickstart hack or something? Is this possible at all?
Two general questions about Fedora 15 before I decide to install. My system typically starts in about 32 seconds, with a distro using normal boot scripts. With Ubuntu and upstart it reaches the gdm login screen quicker in around 18 - 24 seconds. With systemd does anyone get a noticeable difference?
Also, is it possible to use an alternative desktop to gnome3 in Fedora 15. I've tried Gnome3 with Arch and its very slick although not for me, as I like panels and desktop applets. (I should add that my times quoted were mesaured using bootchart with a default set of services, no tweaking).
I've experienced this problem with a distribution upgrade and then subsequently with a clean install of Ubuntu 10.04. Basically half the time that I boot up Ubuntu, it completely locks up with the following messages on screen (Sorry, I had to take a picture since it was locked up)However, when it does boot correctly, it pauses for just a moment after displaying the above text, and then displays a few additional lines, after which it goes on to boot 'normally'.
The next lines show 'plymouth' and 'ureadahead' processes being killed. If these lines show up the boot continues normally, otherwise the boot process hangs at the first picture.It is pretty frustrating to have to perform a hard restart during nearly every other boot, and the hangs seem almost completely random. Sometimes, it happens a couple times in a row, sometimes I can boot normally several times in a row. It also doesn't seem to be related to a cold start, versus a restart.
I'm working on a homeserver which is backupping all the (windows) pc's using cron and mount+rscync every 15 minutes. But between 23:00 PM and 8:00AM or till someone says START I want the PC to Hibernate. Automatically. How can I get this done?
I want to measure the startup time of any GUI app (e.g. firefox) using the time(1) command. However, timing is measured until the app is closed, which has to be done manually by exiting it or clicking X.
How can I get the app to load, terminate immediately and give me the startup time?
I know that uptime prints the time a machine has been up and running, but is there an easier (reliable) way to get the date of the start up than counting down from this output?I tried looking around /proc, but didn't find anything of relevance. There's also a line like this on my dmesg: [ 0.673492] rtc_cmos rtc_cmos: setting system clock to 2011-03-14 14:26:52 UTC (1300112812), but I'm wondering if this method is distribution and kernel version agnostic.
I was wondering if there is any tool or program that stores a copy of frequently used files eg. Binarys, program library's etc. in memory so when they are requested by the OS they load instantly. I'm asking because I have a system with plenty of ram but very slow hard disks. Having programs like opera and java/eclipse load from ram would greatly speed up their start time. Ideally they would be loaded into ram in the background after I log in. Of course all writes made to these files would have to be made to the files on disk for obvious reasons.I don't want the entire OS in ram because it will not fit, just frequently accessed files.
I'm running Ubuntu 10.10 in a computer with a 7200rpm HD, 8Gb Ram, and a Core i7, and the time it takes to load the desktop is insane. There are threads that mention the problem of gnome looking for nonexistent floppy drives and to solve the problem by disabling that in the BIOS (option I don't have). Anyway, besides that problem, by running at startup a gnome-terminal with the iotop command, I noticed that two processes have a huge i/o load on the system: google desktop and ubuntu one. I would like those programs to run as part of the startup process, but be launched after several seconds (to allow the rest of the programs to load). Is there any way I can achieve this? I think there should be a way modifying the commands under startup application, but I cannot find anything that works.
I'm having a problem with a server I just got. The first thing I did when I got it was to format the drive and install ubuntu server, but on booting up I got a black screen and my monitor (samsung syncmaster914v) displayed a bouncing notification with the text "not optimum mode recommended mode: 1280 x 1024 60hz". However at one by using a combination of plugging in the monitor after booting, and pressing random keys on the keyboard I was albe to boot into the server, and was able to access the monitor meu, which told me it was running at 640x480, and 60 hertz. having access to the console I stupidly tried changing the resolution and then rebooted, getting the same black screen.
Me and my mom are sharing a laptop.My mom likes to use Google Chrome. I like to use Opera. Is there any way I can tell the system to start the Opera application , but then instantly minimize it to the tray bar? I want that because : if my mom starts the computer and I'm not around , I still want all the tabs in my Opera to load , without my mom knowing anything about my "malefic" plan ...
I need to analyze the squid access log file for the purpose of my research.For that I need to obtain some correlation among time and the web site visited, machine ip and the web site name etc.It's better if I can draw a graph by analyzing the log file . I found some log analyzers but it didnt provide this.I need to obtain some cor-relation from this log file .
rubyripper starts ripping but then it seems it stops during the analyzation.No action noticable and it does not finish.The last message in the log is: analyzing the files for not matching chunks.I left it for hours and nothing happens.It happens with all formats and with all CD's I have.No error messages or other.With opensuse 11.2. there was no problem at all.
rubyripper version 0.6.0 Kernel: Linux 2.6.34-12-desktop x86_64 Distribution: openSUSE 11.3 (x86_64) KDE: 4.4.4 (KDE 4.4.4) "release 7"
Does anybody have an idea or need any logfiles?A similar problem I have in k3b when ripping audio cd's. It hangs up during analyzation.
Does anybody know an application that sums up the printing from the page_log file and puts the summed amounts in another simple text file? Tried phpPrintAnalyzer but could not figure out how to set it up.
I have a problem for which I imagined a radical solution. Okay, I have a script that uses Alltray's "menu" feature as an interface. It's working greatly, EXCEPT sometimes lxpanel (the panel and tray of LXDE) has a hiccup and restarts, at which point the alltray menu is still running, but the icon is gone from the tray.
I want to somehow test the screen for the existence of a color, a color that is likely only to exist in the icon I've chosen for the chat menu. If the color exists on my screen, then the script can continue to do nothing, but if the color doesn't exist, then that means the alltray icon with my custom icon has disappeared, at which point I should have my bash script close the alltray process and start it again.
How to accomplish the checking of my screen for the existence of a particular color? To simplify, I already have simple screenshot scripts (scrot is good for it), so it would be fine if I could simply analyze an image.