Software :: Ulimit -s 40960 Versus Ulimit Ulimit -s 10240
Jan 20, 2011
I wrote this because i was able to use openmpi to run mpirun on my 12-core workstation rather happily since day 1 I setup the system a few months ago. Yesterday when I tried to run a big job under mpirun, the job crashed rather quickly, the error message was something like mpirun process exited blah blah with signal 11 (Segmentation fault). Interestingly (or annoyingly) a job required less memory ran okay.
Since I never had this problem before, I thought it was the hardware failure. I called my IT guy to explain the problem and he is kind
enough to suggest to put a line
ulimit -s 40960
in my .bashrc. And it works! But I have no clue why mpirun misbehaves out of a sudden, and that ulimit setting solves the problem completely. I would like to learn from this incident.
We are using a hadoop server & we dont want the server memory to overload. So, I have set the ulimit for max memory size and for some time it was working fine but memory was overloading before EOD. I came to know about soft & hard limits and set the hard limit for the maximum memory size in /etc/security/limits.conf file. But, the limits were not shown in ulimit -a command ouput. So, I restarted the server. Then, the limit was shown in the ulimit command output. But, memory is still getting overloaded as you can see the memory used is more than the limit set. Anyone kndly suggest me on this issue.
I'd like to set the max stack size at boot time, preferably as early as possible. I had thought there's a kernel parameter for this, but I cannot find any reference to one.Adding "ulimit -s" to rc.sysinit was not effective; I've grepped /etc/init.d but it is not used subsequently. So what I currently have is my default runlevel set to 1, with no services except udev-post, then in rc.local:
Code:
ulimit -s 1024 telinit 3
3 being my normal runlevel with all necessary services. However, again that ulimit is being ignored, because after login it is still set to 8192. If after login I manually set the limit and change from 3 to 1 and back again (to restart all services), everything is copacetic.
Does anyone know why ulimit -s is ineffective prior to login or how I can set the max stack size at boot time? I'm just noticing that under "upstart" runlevel services are not controlled by rc.d content. Preuming:
#1 My desire to make F14 work without upstart, using ye olde sysinitv which weren't broke so I don't need it fixed, is going to be like p'sing up a rope.
#2 That systemd is replacing upstart so I might as well try to learn about that.
how can I get "rc.local" functionality under systemd?
I need to set ulimit values according to I used Code: ulimit -c unlimited ulimit -s unlimited ulimit -u unlimited But after I log again to my session values are set to default. Now can I set this values so they don't change after log out of after reboot?
I would like to use ulimit -St <some number> to stop a process form running once the process has exceeded the amount of CPU time specified in the ulimt arg.
I have been able to get ulimit -t (or -Ht) to kill my process, but I need the signal to be something other than 9 and other than SIGXCPU.
Is there a way to have ulimit issue a kill -<somthing else> when the cpu time specified has been exceeded?
A strange question, I guess. I'm running processes called from a c main program. The calling is performed (for now) as: FILE * res=popen(ulimit -t 1; prg args); So I can read the stdout of the process as a file and analyze it. The time limit is important for me.
2 questions: 1. How do I get to know if the process terminated on its own or by the ulimit? 2. How do I limit to times that are less than 1 sec (I have many of those).
I know that setrlimit exists, just before I change my whole approach I wanted to see if I can deal with these things from the outside.
Newbie here! Our website CMS is a Tomcat webapp, which runs on a CentOS 5.6 release (Final). The webapp needs a permanent increase of the max. open files value. Currently, the site is "crashing" frequently due to continuous "Too many open files" exceptions that eventually will occur when traffic increases.
This is what I've done to try to increase the max. open files value code...
But still, when I log in (as any user, incl. root), ulimit -n shows 1024, not 16384. Am I missing something here? And, more importantly; Will Tomcat be able to open > 1024 files after my changes mentioned above?
Fedora 12 gcc 4.4.1 I am doing some programming, and my program gave me a stack dump. However, there is no core file for me to examine.
So I did: Code: ulimit -c unlimited and got this error message:
Code: bash: ulimit: core file size: cannot modify limit: Operation not permitted I also tried setting ulimit to 50000 and still got the same error. The results of ulimit -a:
Code: $ ulimit -a core file size (blocks, -c) 0 data seg size (kbytes, -d) unlimited scheduling priority (-e) 0 file size (blocks, -f) unlimited
I just built a computer with a MSI 760GM-E51 Motherboard, AMD Phenom II 955 CPU and 4 GB memory (2 x 2 GB G.Skill DDR3-1333 PC3-10666) I first installed 10.04 LTS 64bit. But I think I am going to go to 11.04. I am just looking for a vote, would you install 32 or 64 bit. I am not a real power user, biggest thing I would do would be video editing. If I am going to play resource demanding games, I would probably boot into XP.
I installed Fedora 13 and use the Gnome desktop. I want to keep my installation as clean as possible and have heard some contrary advice about installing both Gnome and KDE desktops, so I want to stay with just Gnome. In the past I have mixed both and feel that resulted in tons of packages that I probably didn't need and tons of updates all the time. However, some applications seem to be KDE applications and installing them requires installation of many KDE packages. This is a source of confusion for me:
Is there a distinction between Gnome applications and KDE applications? If so, how do you tell the difference? Should one NOT install KDE applications if you are using the Gnome desktop and not interested in installing KDE desktop? Is there a best practice on how to approach which software to install so that you do not create a mix and match mess?
I have a script that basically adds a zypper repo, then proceeds to install and configure FreeNX.To add the repo:
Code:
zypper addrepo Index of /repositories/X11:/RemoteDesktop/openSUSE_11.1 RemoteDesktop
To install FreeNX & it's relevant dependencies:
Code:
zypper install FreeNX
To setup and configure FreeNX:
Code:
nxsetup --install --setup-nomachine-key --clean --purge sed -i 's/AllowUsers idcuser/AllowUsers idcuser nx/' /etc/ssh/sshd_config service sshd reload
[code]....
After completing these steps on version 11, I can immediately open the FreeNX client (windows 7), and connect. On 11.1, at the very end of the FreeNX connect, just after "Dowloading the session information", I get:
Permission denied (publickey,keyboard-interactive).NX> 280 Exiting on signal: 15 I've googled this to death, and tried a bunch of random changes to both ssh and nxserver, but I can't seem to get rid of it. What might have changed from 11 to 11.1 that could cause this behavior change? The NXserver seems to be configured and running identical on both systems.
Are there big disadvantages to building a computer around a motherboard with ATI Radeon graphics instead of Nvidia? I am using an AMD CPU to save money, but all the motherboards AMD recommends use ATI. I have always used Nvidia in the past, And am not sure what the current state of ATI Linux drivers is. I know I would be giving up VDPAU acceleration for video playback, but hopefully the Athlon� II X4 635 processor I am looking at has enough horsepower to handle this on it's own, even for high-def h264.
With the problems I seem to be having I was wondering about data integrity of an optical disc vs a USB drive.How about transfer speed? I assume you can format a usb drive as ext4. I am beginning to wonder why I put 2 new dvdr/rws in my new computer.Is an external hard drive the best solution for backing up files?
I am trying my hand at Bash scripting. I have a file with lots of pinyin, which is the romanized version of Chinese characters and words. A typical entry looks like this,
"7 shuo1 to speak"
Seven is the number of the entry shuo is the pinyin, 1 is the tone mark and is the Chinese character.
What I would like to do is change the format so that it looks more like this:
7 shuo 1
Each field needs to be re-assigned to a variable in an array and then printed to the screen and saved to a file in the same order it was input. This is to prepare the data for another project I am working on in imagemagick.
I have spent all day looking at linux man pages and have very little to show for my efforts. What is the best way to approach this? grep, awk, sed...?
i have some servers behind a server/router/firewall at [URL] that can be accessed using port forwarding. they are working quite well. [URL] gets you to the first server behind the server/router/firewall (the former link above).
i would want to make the server at the latter link accessible with its own domain name ie [URL]without having to add :<port> to the end of [URL] because ":" is disallowed for aliasing in the a and cname records at network solutions (my dns?). is port forwarding the best way to reach these servers behind my firewall? is it possible to assign them their own domain name? perhaps some method other than port forwarding should be used?
trying to do a multiboot, just for fun Now I installed Foresight Linux, which was not such a good.Foresight is based on rpath and uses Conary as update system Now Conary destroyed all other linux systems installed on the other partitions. Now I found that there is something such as hide and unmount but have read several pages full of it but still have no idea what is the difference between the two and more important, how to use this as most explanations seem to complicated How can I hide partitions for a booted linux operating system so it is unable to see it, use it or even mount it when it tries
I am integrating my Unix box to the Windows AD using PAM_LDAP and Kerberos enabled. I was wondering, since Kerberos is enabled is there any point to enable SSL on my LDAP.conf? My understanding is that since Kerberos is enabled, therefore the username/password is sent securely there isn't any benefit of enabling SSL on the LDAP.conf? It's one of or another.
I am leading a project at work that will require at least one new server. There will be a development server and a production server, which changes from development will be rolled onto. Unfortunately, I am more of a web programmer than a Linux guru, and I really don't know whether or not it is better to have two physical servers or two virtual servers on one machine. I don't think there will be a huge toll on the machine, as there will probably be around 1000 total users and less than 100 on at any given time. We are also able to spend quite a bit on the server, so I'm sure one could handle it, but I just wanted to check and see what the advantages and disadvantages would be in this situation.
It is my understanding that building from source is slightly faster since the created binary will be specific to the current kernel/system. This seems to have the drawback of when the kernel is upgraded the source compiled package will no longer work...but this is not always the case. So what kernel upgrades will cause a package from source not to function? Side bar to that, I have a package that is only available in source format that I need to install on numerous hosts. It seems that the upfront work of creating my own RPM for this package is worth it to make managing the package more efficient, and I think I have two choices: src-rpm and standard binary rpm. If I build a src-rpm and I distribute that for installation will a kernel upgrade 'break' the package since it is actually being compiled for a specific kernel at the time of install? So far I have found a lot of information, and have spent time reading the rpm docs, etc, but I don't feel like I have a good grasp of what the specific portability differences are between source, src-rpm, and rpm.
Which do you prefer and why? Also, I remember someone telling me once that while using space indentation use either 4 spaces or 2 spaces. Don't use 3 spaces. Anyone know why they would say not to use 3 spaces?
I am confused with the concept of Kerberos and LDAP SSL. I am in the midst of integrating my Unix box with the Active Directory hence the use of PAM_LDAP method. I understand that since it's non-secure transmission hence We use Kerberos to authenticate. If we already used kerberos to authenticate i.e. it means that the username/password is not transmitted in clear text. Why we still need LDAP SSL? What is the benefit?
I recently made an Ubuntu liveCD for a family member, and tested it myself first. It impressed me, so I might possibly be interested in switching distros; but I'm not very familiar with Gnome(GNOME?), so I might try Kubuntu instead. Am I correct in understanding that the only difference is Kubuntu uses KDE (hence the K)? If there are other differences, does Kubuntu share the limited user control that I've heard Ubuntu has, such as locking the root account by default?
