Thrashing on Ubuntu seems to make it almost impossible to interact with the system. It can be minutes before any interaction has an effect, including remote connections. This means that if a program either misbehaves or simply needs more memory than available RAM, it might be hard or impossible to stop it, either locally or remotely. This can be both an annoyance and a security threat (since a process without elevated privileges can effectively hang the system).
It is not hard to make a system go into thrashing, especially if it is low on memory (that's probably true in general, not only of Ubuntu). On my ASUS netbook running Ubuntu 10.04, with only 1GB RAM, thrashing can occur as easily as running both Chromium (which is a bit of a memory hog) and the Resynthesizer plugin in the GIMP at the same time. Running these programs plus another memory-intensive program like Mathematica can generate thrashing even when 2GB or 4GB of physical memory are available.
I am including a short C++ program that allocates and accesses a large amount of memory, guaranteeing thrashing will occur on any system. Using this or any other memory-intensive program, the steps required to reproduce the condition I described are
1. Start one or more memory-intensive programs.
2. As RAM is filled, paging will start, and if the programs try to access the memory that has been swapped out, thrashing occurs.
What happens?
- Interactivity with the system drops to almost zero. Mouse barely moves, keyboard interaction has huge delays (tens of seconds), starting a terminal or switching to one if one is already open can take minutes, as is the case with remote (e.g. SSH) connections.
What I would expect/want to happen?
- The system should keep interactivity levels high at all times. While I'm not at all an expert on this, I would think this could be achieved by either not allowing paging out of essential user-interface elements, or more generally by giving processes that generate a lot of page faults comparatively lower priority than other processes, especially processes that are just starting, or are part of the user interface.
To use the included program, compile with
g++ -o bug bug.cc
and run with
./bug <amount of memory>
where <amount of memory> is the amount of memory in MB to be allocated. One can run several instances of the program at the same time, to compete for memory. One can use top, free or the System Monitor to check when RAM is completely filled, and thrashing starts.
PS: This seems to be a long-standing issue with Linux, it's not limited to the current version of Ubuntu.
ProblemType: Bug
DistroRelease: Ubuntu 10.04
Package: linux-image-2.6.32-24-generic 2.6.32-24.39
Regression: No
Reproducible: Yes
ProcVersionSignature: Ubuntu 2.6.32-24.38-generic 2.6.32.15+drm33.5
Uname: Linux 2.6.32-24-generic x86_64
NonfreeKernelModules: nvidia
AlsaVersion: Advanced Linux Sound Architecture Driver Version 1.0.21.
Architecture: amd64
AudioDevicesInUse:
USER PID ACCESS COMMAND
/dev/snd/controlC0: tibi 1596 F.... pulseaudio
/dev/snd/pcmC0D0p: tibi 1596 F...m pulseaudio
/dev/snd/controlC1: tibi 1596 F.... pulseaudio
CRDA: Error: [Errno 2] No such file or directory
Card0.Amixer.info:
Card hw:0 'NVidia'/'HDA NVidia at 0xfe020000 irq 21'
Mixer name : 'Nvidia MCP78 HDMI'
Components : 'HDA:10ec0888,10250153,00100202 HDA:10de0002,10de0101,00100000'
Controls : 37
Simple ctrls : 20
Card1.Amixer.info:
Card hw:1 'U0x46d0x8da'/'USB Device 0x46d:0x8da at usb-0000:00:04.0-2, full speed'
Mixer name : 'USB Mixer'
Components : 'USB046d:08da'
Controls : 3
Simple ctrls : 2
Date: Wed Aug 18 13:01:30 2010
HibernationDevice: RESUME=UUID=9464cfb9-e39a-46ab-bf3a-01a7f2194ab1
InstallationMedia: Ubuntu 10.04 LTS "Lucid Lynx" - Release amd64 (20100429)
IwConfig:
lo no wireless extensions.
eth0 no wireless extensions.
MachineType: eMachines EL1210-09
ProcCmdLine: BOOT_IMAGE=/boot/vmlinuz-2.6.32-24-generic root=UUID=93f3d657-74ac-4853-acab-704f74ab4cd0 ro quiet splash
ProcEnviron:
PATH=(custom, user)
LANG=en_US.utf8
SHELL=/bin/bash
RelatedPackageVersions: linux-firmware 1.34.1
RfKill:
SourcePackage: linux
dmi.bios.date: 09/23/2008
dmi.bios.vendor: Phoenix Technologies, LTD
dmi.bios.version: R01-A0
dmi.board.name: WMCP78M
dmi.board.vendor: eMachines
dmi.chassis.asset.tag: 0000000000000000000000
dmi.chassis.type: 3
dmi.chassis.vendor: eMachines
dmi.modalias: dmi:bvnPhoenixTechnologies,LTD:bvrR01-A0:bd09/23/2008:svneMachines:pnEL1210-09:pvrR01-A0:rvneMachines:rnWMCP78M:rvr:cvneMachines:ct3:cvr:
dmi.product.name: EL1210-09
dmi.product.version: R01-A0
dmi.sys.vendor: eMachines
This is an attempt at bringing sanity to bug #7372. Please only comment here is you are experiencing high I/O wait times and interactvity on reasonable workloads.
Latest working kernel version: 2.6.18?
Problem Description:
I/O operations on large files tend to produce extremely high iowait times and poor system I/O performance (degraded interactivity). This behavior can be seen to varying degrees in tasks such as,
- Backing up /home (40GB with numerous large files) with diffbackup to external USB hard drive
- Moving messages between large maildirs
- updatedb
- Upgrading large numbers of packages with rpm
Steps to reproduce:
The best synthetic reproduction case I have found is,
$ dd if=/dev/zero of=/tmp/test bs=1M count=1M
During this copy, IO wait times are very high (70-80%) with extremely degraded interactivity although throughput averages about 29MB/s (about the disk's capacity I think). Even starting a new shell takes minutes, especially after letting the machine copy for a while without being actively used. Could this mean it's a caching issue?