42

I'm looking for a way to limit a processes disk io to a set speed limit. Ideally the program would work similar to this:

$ limitio --pid 32423 --write-limit 1M

Limiting process 32423 to 1 megabyte per second hard drive writing speed.

Improve this question
0

4 Answers 4

Reset to default
44

That is certainly not trivial task that can't be done in userspace. Fortunately, it is possible to do on Linux, using cgroup mechanizm and its blkio controller.

Setting up cgroup is somehow distribution specific as it may already be mounted or even used somewhere. Here's general idea, however (assuming you have proper kernel configuration):

mount -t tmpfs cgroup_root /sys/fs/cgroup
mkdir -p /sys/fs/cgroup/blkio
mount -t cgroup -o blkio none /sys/fs/cgroup/blkio

Now that you have blkio controller set, you can use it:

mkdir -p /sys/fs/cgroup/blkio/limit1M/
echo "X:Y  1048576" > /sys/fs/cgroup/blkio/limit1M/blkio.throttle.write_bps_device

Now you have a cgroup limit1M that limits write speed on device with major/minor numbers X:Y to 1MB/s. As you can see, this limit is per device. All you have to do now is to put some process inside of that group and it should be limited:

echo $PID > /sys/fs/cgroup/blkio/limit1M/tasks

I don't know if/how this can be done on other operating systems.

Improve this answer
2
  • 4
    Please note that you first need to set the device wide total policy in order to create sub-groups by i.e. echo "X:Y 1073741824" > /sys/fs/cgroup/blkio/blkio.throttle.write_bps_device and you obtain X and Y by i.e. ls -l /dev/sda
    – dothebart
    Commented Apr 20, 2018 at 11:32
  • cgroup v1 doesn't support following cache's writeback. That means that testing a simple dd command writing to a standard file in a filesystem (rather than to a device) will probably not show any limitation. Not using any writeback will work: dd ... oflag=direct will work as expected. cgroup v2 can handle this if the filesystem supports it. Details: Using cgroups to limit I/O
    – A.B
    Commented Nov 1, 2019 at 10:25
24

ionice from the util-linux does something similar to what you want.

It doesn't set absolute IO limits, it sets IO priority and 'niceness' - similar to what nice does for a process' CPU priority.

From the man page:

ionice - set or get process I/O scheduling class and priority

DESCRIPTION
This  program  sets or gets the I/O scheduling class and priority for a
program.  If no arguments or just -p is given, ionice  will  query  the
current I/O scheduling class and priority for that process.

When  command  is  given,  ionice  will run this command with the given
arguments.  If no class is specified, then  command  will  be  executed
with the "best-effort" scheduling class.  The default priority level is 4.
Improve this answer
1
  • 11
    One should bear in mind that it works only with a CFQ I/O scheduler. Many modern systems have deadline one enabled by default.
    – Highstaker
    Commented May 3, 2016 at 8:44
14

The answer of fche is a very good hint, thanks for that, although it dosn't really solve the problem cause the question was to limit a process to a specific bandwidth.

I would suggest something like this:

systemd-run -p "IOWriteBandwidthMax=/dev/sdX 1M" updatedb

or the deprecated version:

systemd-run -p "BlockIOWriteBandwidth=/dev/sdX 1M" updatedb

Still this does not fit the question cause it cannot be used for an already running process but maybe it's helpfull in some other cases.

Links:

Improve this answer
13

systemd provides a wrapper for cgroup-manipulated invocations of processes. From the systemd-run(1) man page:

The following command invokes the updatedb(8) tool, but lowers the block IO weight for it to 10. See systemd.resource-control(5) for more information on the BlockIOWeight= property.
systemd-run -p BlockIOWeight=10 updatedb

Consider using the --scope option to make systemd-run run the program in the foreground.

Improve this answer

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .