Re: [systemd-devel] Problems with systemd-coredump
On 02/18/2014 11:05 AM, Thomas Bächler wrote: Am 17.02.2014 21:27, schrieb Manuel Reimer: As soon as a bigger coredump (about 500 MB) is to be stored, the whole system slows down significantly. Seems like storing such big amounts of data takes pretty long and is a very CPU hungry process... I completely agree. Since the kernel ignores the maximum coredump size when core_pattern is used, a significant amount of time passes whenever a larger process crashes, with no benefit (since the dump never gets saved anywhere). This is extremely annoying if processes with sizes in the tens or hundreds of gigabytes crash, which sadly happened to me quite a few times recently. If this feature is broken by design, why is it still enabled by default on Arch Linux? systemd-coredump makes it nearly impossible to debug bigger processes and it took me quite some time to figure out how to get coredumps placed to /var/tmp so I can use them to find out why my process has crashed. Yours Manuel ___ systemd-devel mailing list systemd-devel@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/systemd-devel
Re: [systemd-devel] Problems with systemd-coredump
Am 17.02.2014 21:27, schrieb Manuel Reimer: As soon as a bigger coredump (about 500 MB) is to be stored, the whole system slows down significantly. Seems like storing such big amounts of data takes pretty long and is a very CPU hungry process... I completely agree. Since the kernel ignores the maximum coredump size when core_pattern is used, a significant amount of time passes whenever a larger process crashes, with no benefit (since the dump never gets saved anywhere). This is extremely annoying if processes with sizes in the tens or hundreds of gigabytes crash, which sadly happened to me quite a few times recently. signature.asc Description: OpenPGP digital signature ___ systemd-devel mailing list systemd-devel@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/systemd-devel
Re: [systemd-devel] Problems with systemd-coredump
On Tue, Feb 18, 2014 at 11:05 AM, Thomas Bächler tho...@archlinux.org wrote: Am 17.02.2014 21:27, schrieb Manuel Reimer: As soon as a bigger coredump (about 500 MB) is to be stored, the whole system slows down significantly. Seems like storing such big amounts of data takes pretty long and is a very CPU hungry process... I completely agree. Since the kernel ignores the maximum coredump size when core_pattern is used, a significant amount of time passes whenever a larger process crashes, with no benefit (since the dump never gets saved anywhere). This is extremely annoying if processes with sizes in the tens or hundreds of gigabytes crash, which sadly happened to me quite a few times recently. It's an incomplete and rather fragile solution the way it works today. We cannot really *malloc()* the memory for a core dump, it's *pipe* from the kernel for a reason. It can be as large as the available RAM, that's why it's limited to the current maximum size, and therefore also limited in its usefulness. It really always needs to be extracted to be a minidump to store away. There are no other sensible options when things should end up in the journal. Kay ___ systemd-devel mailing list systemd-devel@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/systemd-devel
[systemd-devel] Problems with systemd-coredump
Hello, if a bigger application crashes with coredump, then systemd-coredump seems to have a few problems with that. At first, there is the 767 MB limitation which just drops all bigger coredumps. But even below this limit it seems to be impossible to store coredumps. I did a few tries and found out that, with default configuration, the limit seems to be at about 130 MB. Bigger coredumps are just dropped and I cannot find any errors logged to somewhere. It seems to be possible to work around this problem by increasing SystemMaxFileSize to 1000M. With this configuration change, bigger coredumps seem to be possible, but this causes another problem. As soon as a bigger coredump (about 500 MB) is to be stored, the whole system slows down significantly. Seems like storing such big amounts of data takes pretty long and is a very CPU hungry process... Can someone please give some informations on this? Maybe it's a bad idea to store such big amounts of data in the journal? If so, what's the solution? Will journald get improvements in this area? Thank you very much in advance. Greetings, Manuel ___ systemd-devel mailing list systemd-devel@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/systemd-devel
Re: [systemd-devel] Problems with systemd-coredump
On Mon, Feb 17, 2014 at 9:27 PM, Manuel Reimer manuel.s...@nurfuerspam.de wrote: Hello, if a bigger application crashes with coredump, then systemd-coredump seems to have a few problems with that. At first, there is the 767 MB limitation which just drops all bigger coredumps. But even below this limit it seems to be impossible to store coredumps. I did a few tries and found out that, with default configuration, the limit seems to be at about 130 MB. Bigger coredumps are just dropped and I cannot find any errors logged to somewhere. It seems to be possible to work around this problem by increasing SystemMaxFileSize to 1000M. With this configuration change, bigger coredumps seem to be possible, but this causes another problem. As soon as a bigger coredump (about 500 MB) is to be stored, the whole system slows down significantly. Seems like storing such big amounts of data takes pretty long and is a very CPU hungry process... Can someone please give some informations on this? Maybe it's a bad idea to store such big amounts of data in the journal? If so, what's the solution? Will journald get improvements in this area? Thank you very much in advance. Greetings, Manuel I wish there was a good way to install a system debugger which could inspect the process and its memory at the time of the crash and generate a short textual report, like libSegFault, or a minidump, like breakpad. Either hopefully small enough to just chuck into the journal. core_pattern requires funneling the entire process memory through a pipe and making a copy of it. LD_PRELOAD seems terribly brittle and doesn't work on statically linked binaries. ___ systemd-devel mailing list systemd-devel@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/systemd-devel
Re: [systemd-devel] Problems with systemd-coredump
On Mon, Feb 17, 2014 at 9:43 PM, Jan Alexander Steffens jan.steff...@gmail.com wrote: On Mon, Feb 17, 2014 at 9:27 PM, Manuel Reimer manuel.s...@nurfuerspam.de wrote: Hello, if a bigger application crashes with coredump, then systemd-coredump seems to have a few problems with that. At first, there is the 767 MB limitation which just drops all bigger coredumps. But even below this limit it seems to be impossible to store coredumps. I did a few tries and found out that, with default configuration, the limit seems to be at about 130 MB. Bigger coredumps are just dropped and I cannot find any errors logged to somewhere. It seems to be possible to work around this problem by increasing SystemMaxFileSize to 1000M. With this configuration change, bigger coredumps seem to be possible, but this causes another problem. As soon as a bigger coredump (about 500 MB) is to be stored, the whole system slows down significantly. Seems like storing such big amounts of data takes pretty long and is a very CPU hungry process... Can someone please give some informations on this? Maybe it's a bad idea to store such big amounts of data in the journal? If so, what's the solution? Will journald get improvements in this area? I wish there was a good way to install a system debugger which could inspect the process and its memory at the time of the crash and generate a short textual report, like libSegFault, or a minidump, like breakpad. Either hopefully small enough to just chuck into the journal. That is the plan, and someone just needs to finish it: http://cgit.freedesktop.org/libminidump Kay ___ systemd-devel mailing list systemd-devel@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/systemd-devel