Summary: | CPU at ~100% and memory usage very high when downloading a large file (>4GB) | ||
---|---|---|---|
Product: | [Applications] kget | Reporter: | Praveen Srinivasan <popapa> |
Component: | Multisegkio | Assignee: | KGet authors <kget> |
Status: | RESOLVED WAITINGFORINFO | ||
Severity: | normal | CC: | andresbajotierra, axel.braun, drotos, mat69, sgh, tim |
Priority: | NOR | ||
Version: | unspecified | ||
Target Milestone: | --- | ||
Platform: | Ubuntu | ||
OS: | Linux | ||
Latest Commit: | Version Fixed In: | ||
Sentry Crash Report: |
Description
Praveen Srinivasan
2008-04-01 21:19:57 UTC
I assume you are using KGet from KDE 4.0.2, right? Yes, I am. On 1 Apr 2008 19:43:04 -0000, Urs Wolfer <uwolfer@kde.org> wrote: [bugs.kde.org quoted mail] Don't know whether it's the same bug, but I also get high CPU usage (not 100%, but high) using kde4's kget, though it doesn't depend on the size of file, but on download speed - e.g. when downloading at 10Mbit/s. However, cpu load is coming from kget and kio_file processes. I don't have that kind of problem with old, kde3 kget. For comparison, wget has almost 0% CPU usage when downloading the same files on my machine. I didn't test with smaller files at high speed, so that may well be it. -Praveen On Wed, Apr 2, 2008 at 9:48 AM, Kolik <kolik777@gmail.com> wrote: [bugs.kde.org quoted mail] Mmh, 10 MBit/s is very much =) Anyway, that could be, cause of a function in KDELibs which takes uch cpu =( We have to optimize that for 4.1 :) Lukas In my case, it was downloading at ~2MB/sec (megabytes/sec) (over 100MBit link). Th problem still exists with kde 4.0.4 I am downloading with 1.6MB/s (Megabytes per second) And i have also ~73% us-load kio_file uses ~60% and kget ~14% Can someone test it with the latest KGet from 4.1rc1? LUkas Version 2.1.0 Using KDE 4.1.00 (KDE 4.1.0) (KDEmod) in ArchLinux i686: After 8 minutes of downloading these two files, these values were constant: "ps uax" output: dario 5239 5.8 2.7 72344 28660 ? R 13:59 0:24 /usr/bin/kget -caption KGet -icon kget While downloading (copying) a file from another computer in my network (using my wifi connection) at 1.4MB per second, these values were obtained ("ps uax"): At start: dario 5366 2.4 2.4 78424 25580 ? S 14:12 0:01 /usr/bin/kget -caption KGet -icon kget At end: dario 5366 5.8 2.5 78424 25740 ? R 14:12 0:05 /usr/bin/kget -caption KGet -icon kget (the CPU usage was slowly increasing during the transfer) After the download KGet CPU usage is slowly decreasing I still have the bug in KDE 4.1; kget was maxing-out the CPU (there were other things running, so it couldn't get to 100%, but most likely would have). I experience this problem mainly when using the torrent plugin. For smaller files (~200MB) everything works fine, but for large files (>=1GB) the CPU usage of KGet is nearly 100% and the memory consumption is constantly growing (after 6 hours it's around 1.5GB, and still growing). This behavior does not depend on the download/upload speed for me, the resource usage does not decrease even if I pause the download. System: Debian amd64, unstable/experimental KGet version: 4.1.2-1 (Debian versioning), 2.1.2 ('about' dialog) So I think the Torrent-Plugin-Issue is more the fault of libbtcore, which is handling all that stuff for us... can u test if ktorrent has the same problem? Lukas I've installed ktorrent (KDE 4.1.2, ktorrent 3.1.4, Debian unstable/experimental), and I downloaded a 2GB file with it. The CPU- and memory usage of ktorrent was low all the time (max 5-6% CPU, ~100MB mem; the download took about 7 hours). Interestingly after my last comment, when I restarted KGet to get my resources back, the resource consumption stayed normal. However earlier I've experienced the opposite, too. From my point of view I can't decide if this is a problem with libbtcore, and I've only seen it using KGet because I've only tested ktorrent once; or it's due to a bug in KGet. From now on I will use both tools for torrent downloading and I'll keep an eye on the resource usage, and I'll report if I've found anything useful. Ok great! Hope you find something :) Lukas I am having this exact problem when downloading a 21G file over http. Top looks like this: PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 10090 sgh 20 0 79516 25m 19m R 59 2.9 11:12.08 kget 10879 sgh 20 0 103m 73m 3880 S 23 8.3 1:31.59 kio_file 10880 sgh 20 0 39132 9684 6452 S 16 1.1 1:10.36 kio_http Well ... kio_file also has a huge memory-consumption. That is just crazy - 149 Megs just for fetching a file! PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 10090 sgh 20 0 80020 26m 19m R 56 2.9 15:41.85 kget 10879 sgh 20 0 180m 149m 3880 S 23 16.9 3:15.56 kio_file 10880 sgh 20 0 39132 9684 6452 S 17 1.1 2:23.41 kio_http With KGet 2.2.0 (KDE 4.2) from Ubuntu PPA packages I still have the extremely high cpu usage problem with KGet. I've had this same problem with KGet from every version I've tried. CPU usage is always high even when downloading over a slow 10 Mbit / sec link. Over gigabit lan it's almost impossible to max out the transfer rate because CPU usage hits 100%. With wget on the same files the cpu usage is at roughly zero percent. It's a shame because KGet could be so good! It seems silly that a file downloader would bring a modern cpu to its knees. I should report that while downloading at approx 1MB/sec I see the following cpu usage: I should report that while downloading at approx 1MB/sec I see the following cpu usage: - kget: 37% - plasma: 9% - kio_file: 6% - kio_http: 6% With KGet disabled the plasma usage goes way down (I do not have any kget widgets active). (In reply to comment #13) > > From now on I will use both tools for torrent downloading and I'll keep an eye > on the resource usage, and I'll report if I've found anything useful. > In KDE >=4.1.3 kget works fine for me even when a really huge file (8GB) is downloaded via torrent (low CPU, reasonable memory usage), so my problem is solved. It seems that this was a different issue, it was not related to this bug. btw please retest in latest trunk... there might've been some improvements. Lukas *** Bug 213050 has been marked as a duplicate of this bug. *** Please reopen if this problem still exists in KDE 4.4. |