add blkio.weight support by hqhq · Pull Request #11802 · moby/moby

hqhq · 2015-03-26T07:25:40Z

We can use this to control block IO weight of a container.

Signed-off-by: Qiang Huang [email protected]

moxiegirl · 2015-03-26T16:55:05Z

A better structure for this is: https://gist.github.com/fb433f62445d80cacf48.git

In tech docs, you typically don't want to talk about features that "might" be coming. Creates an anticipation in the reader that may or may not ever be resolved. A feature either is or is not available.
--->

Block IO bandwidth (Blkio) constraint

By default, all containers get the same proportion of block IO bandwidth
(blkio). This proportion is 500. To modify this proportion, change the
container's blkio weight relative to the weighting of all other running
containers using the --blkio-weight flag.

The --blkio-weight flag can set the weighting to a value between 10 to 1000.
For example, the commands below create two containers with different blkio
weight:

$ sudo docker run -ti --name c1 --blkio-weight 300 ubuntu:14.04 /bin/bash $ sudo docker run -ti --name c2 --blkio-weight 600 ubuntu:14.04 /bin/bash

If you do block IO in the two containers at the same time, by, for example:

time dd if=/mnt/zerofile of=test.out bs=1M count=1024 oflag=direct

You'll find that the proportion of time is the same as the proportion of blkio
weights of the two containers.

Note: The blkio weight setting is only available for direct IO. Buffered IO
is not currently supported.

moxiegirl · 2015-03-26T16:56:41Z

Thanks for the contribution!

LK4D4 · 2015-03-26T18:07:56Z

This will require docs bump.

hqhq · 2015-03-27T01:36:07Z

@moxiegirl Thanks for your suggestion, updated.

moxiegirl · 2015-03-29T20:52:41Z

Thanks @hqhq --- much appreciated. I probably jumped the gun a bit commenting docs. This needs a code review.

vieux · 2015-04-15T21:20:01Z

Could you possibly start 2 containers, one with high weight and the other with low weight and make sure the high weight finished first ?

That would be a bit difficult, but doable, the thing is I think it's not necessary.
All Docker needs to be sure is that we write the right value into cgroup files, the functionality is guarantied by cgroup, which is kernel's job. And we don't need to test kernel functionality in Docker. Otherwise, all other resource configs need test to make sure if these resource limit really work. WDYT?

Can we also have test cases for invalid values or --blkio-weight?

vieux · 2015-04-15T21:25:07Z

code LGTM besides a few nits.

Can you also move the remote api doc change from 1.18 to 1.19 thanks!

hqhq · 2015-04-16T07:19:18Z

Force update mistakenly close the PR, and I can't reopen it, somebody help?

hqhq · 2015-04-16T07:29:35Z

Sorry, my mistake, reopen it.

hqhq · 2015-04-16T07:30:13Z

Updated and rebased.

icecrime · 2015-04-16T16:29:11Z

No \n at the end of the error message please (leave it to the printing code, if any)!

vbatts · 2015-04-17T03:07:33Z

I am not familiar and will have to check.
On Apr 16, 2015 12:33 PM, "Arnaud Porterie" [email protected]
wrote:

A thing that confuses me about the parameter value is the following:

kernel.org
https://www.kernel.org/doc/Documentation/cgroups/blkio-controller.txt
says:

Currently allowed range of weights is from 10 to 1000.
-
RHEL docs
https://access.redhat.com/documentation/en-US/Red_Hat_Enterprise_Linux/6/html/Resource_Management_Guide/ch-Subsystems_and_Tunable_Parameters.html#blkio-weight
say:

[...] in the range 100 to 1000

Ping @vbatts https://github.com/vbatts: wdyt?

—
Reply to this email directly or view it on GitHub
#11802 (comment).

hqhq · 2015-04-17T03:40:54Z

@icecrime Updated, thanks.

vbatts · 2015-04-17T13:53:02Z

after double checking, RHEL6 did not get that commit. RHEL6 minimum is 100.

icecrime · 2015-04-17T15:07:16Z

Thanks for checking @vbatts! That's embarrassing, but I can't see us handling this at Docker level so I think we're just gonna go with the 10 to 1000 range, and docker run --blkio-weight=20 will most likely fail on RHEL.

LGTM 👍 Still good to you @vieux?

vbatts · 2015-04-17T15:18:01Z

@icecrime fairly narrow affected case. RHEL6 users, that call docker run with --blkio-weight and the value provided is 10-99, they'll get an invalid argument

vbatts · 2015-04-17T17:27:35Z

We don't have to hold up this PR on RHEL6. I've opened a request for torvalds/linux@df457f8 in this bz https://bugzilla.redhat.com/show_bug.cgi?id=1212931 (permissions of this bz may change over time).

I'll look to add docs (... somewhere) about the narrow case.

icecrime · 2015-04-17T22:17:04Z

Thanks @vbatts, sounds good!

calavera · 2015-04-29T20:55:25Z

Code LGTM. @hqhq this needs a rebase, but I'm moving it to docs review so we can get it one step closer. I'll still check the code again before landing it in master.

thaJeztah · 2015-05-05T01:55:45Z

between 10 and 1000

(To be fully correct, "between 9 and 1001", but I don't think that will help anyone)

Looks like you missed this one :)

hqhq · 2015-05-07T02:32:01Z

@thaJeztah Thanks for your review, updated and rebased.

thaJeztah · 2015-05-07T03:55:44Z

Thanks, @hqhq. Looks like you missed one (#11802 (comment)).

LGTM (after that one is fixed)

cc @fredlf @moxiegirl @jamtur01 @SvenDowideit

(why doesn't @docker/docs-owners work?)

We can use this to control block IO weight of a container. Signed-off-by: Qiang Huang <[email protected]>

hqhq · 2015-05-07T04:49:17Z

@thaJeztah Done, thanks.

add blkio.weight support

GordonTheTurtle added status/0-triage dco/yes labels Mar 26, 2015

hqhq mentioned this pull request Mar 26, 2015

enable cgroups memory.oom_control #11034

Merged

hqhq force-pushed the hq_add_blkio_weight branch from b500e5a to 2cda03c Compare March 26, 2015 08:00

moxiegirl reviewed Mar 26, 2015
View reviewed changes

LK4D4 added status/2-code-review and removed status/0-triage labels Mar 26, 2015

hqhq force-pushed the hq_add_blkio_weight branch from 2cda03c to 713dbce Compare March 27, 2015 01:35

icecrime removed the dco/yes label Mar 30, 2015

jessfraz removed the status/0-triage label Apr 3, 2015

vieux reviewed Apr 15, 2015
View reviewed changes

hqhq closed this Apr 16, 2015

hqhq force-pushed the hq_add_blkio_weight branch from 713dbce to ad5d655 Compare April 16, 2015 07:16

GordonTheTurtle added dco/yes and removed dco/no labels Apr 16, 2015

hqhq reopened this Apr 16, 2015

GordonTheTurtle added dco/yes and removed dco/no labels Apr 16, 2015

icecrime reviewed Apr 16, 2015
View reviewed changes

hqhq force-pushed the hq_add_blkio_weight branch from 3634705 to 8c252aa Compare April 17, 2015 03:38

GordonTheTurtle added dco/yes and removed dco/no labels Apr 17, 2015

icecrime removed the dco/yes label Apr 23, 2015

calavera added status/3-docs-review and removed status/2-code-review labels Apr 29, 2015

calavera self-assigned this Apr 29, 2015

thaJeztah reviewed May 5, 2015
View reviewed changes

hqhq force-pushed the hq_add_blkio_weight branch from 8c252aa to fa7f557 Compare May 7, 2015 02:29

add blkio.weight support

f133f11

We can use this to control block IO weight of a container. Signed-off-by: Qiang Huang <[email protected]>

hqhq force-pushed the hq_add_blkio_weight branch from fa7f557 to f133f11 Compare May 7, 2015 04:48

jessfraz pushed a commit that referenced this pull request May 8, 2015

Merge pull request #11802 from hqhq/hq_add_blkio_weight

de32f5c

add blkio.weight support

jessfraz merged commit de32f5c into moby:master May 8, 2015

hqhq deleted the hq_add_blkio_weight branch May 8, 2015 01:50

Conversation

hqhq commented Mar 26, 2015

Uh oh!

moxiegirl Mar 26, 2015

Choose a reason for hiding this comment

Block IO bandwidth (Blkio) constraint

Uh oh!

moxiegirl commented Mar 26, 2015

Uh oh!

LK4D4 commented Mar 26, 2015

Uh oh!

hqhq commented Mar 27, 2015

Uh oh!

moxiegirl commented Mar 29, 2015

Uh oh!

vieux Apr 15, 2015

Choose a reason for hiding this comment

Uh oh!

hqhq Apr 16, 2015

Choose a reason for hiding this comment

Uh oh!

icecrime Apr 16, 2015

Choose a reason for hiding this comment

Uh oh!

vieux commented Apr 15, 2015

Uh oh!

hqhq commented Apr 16, 2015

Uh oh!

hqhq commented Apr 16, 2015

Uh oh!

hqhq commented Apr 16, 2015

Uh oh!

icecrime Apr 16, 2015

Choose a reason for hiding this comment

Uh oh!

vbatts commented Apr 17, 2015

Uh oh!

hqhq commented Apr 17, 2015

Uh oh!

vbatts commented Apr 17, 2015

Uh oh!

icecrime commented Apr 17, 2015

Uh oh!

vbatts commented Apr 17, 2015

Uh oh!

vbatts commented Apr 17, 2015

Uh oh!

icecrime commented Apr 17, 2015

Uh oh!

calavera commented Apr 29, 2015

Uh oh!

thaJeztah May 5, 2015

Choose a reason for hiding this comment

Uh oh!

thaJeztah May 7, 2015

Choose a reason for hiding this comment

Uh oh!

hqhq commented May 7, 2015

Uh oh!

thaJeztah commented May 7, 2015

Uh oh!

hqhq commented May 7, 2015

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

10 participants