[Top] [All Lists]

Re: Enforcing quota for root user

To: Dave Chinner <david@xxxxxxxxxxxxx>
Subject: Re: Enforcing quota for root user
From: Chris Murphy <lists@xxxxxxxxxxxxxxxxx>
Date: Tue, 11 Nov 2014 16:35:31 -0700
Cc: Stan Hoeppner <stan@xxxxxxxxxxxxxxxxx>, "xfs@xxxxxxxxxxx" <xfs@xxxxxxxxxxx>
Delivered-to: xfs@xxxxxxxxxxx
In-reply-to: <20141111223203.GS28565@dastard>
References: <67EDE14C-CCFB-4B42-8B78-FC6C813CD19A@xxxxxxx> <5461B333.8050609@xxxxxxxxxxxxxxxxx> <F7ACFF71-460D-40E7-999C-BFC8D232EA81@xxxxxxx> <54622CE8.7050508@xxxxxxxxxxxxxxxxx> <631A9090-F9FE-4B8E-88CD-1B30CAFC65B7@xxxxxxxxxxxxxxxxx> <20141111223203.GS28565@dastard>

On Nov 11, 2014, at 3:32 PM, Dave Chinner <david@xxxxxxxxxxxxx> wrote:

On Tue, Nov 11, 2014 at 11:00:25AM -0700, Chris Murphy wrote:

On Nov 11, 2014, at 8:36 AM, Stan Hoeppner <stan@xxxxxxxxxxxxxxxxx> wrote:

On 11/11/2014 04:22 AM, Cyril Scetbon wrote:
That's what I've read first, but someone showed me a sample where it works. He just told me he was using project quota. However, does it make sense ?
I've also read somewhere else that quota is never enforced for root user (id,gid=0) that's why I was testing it ....

No, it doesn't make sense.  Why would you want to enforce quotas for root?

A week ago I tried this and project quotas appear to apply to root.

By intent and design. Project quotas are not a user/group based
quota and so there is no exemption for any user.

It’s what I’d expect. I considered the documentation to be slightly misleading where is says soft and hard limits are never applied to the root user. The project quota does have a soft and hard limit. The soft limit, seemed not to apply to the root user - at least there was no soft limit warning anywhere when it was busted. But the hard limit definitely applied.

[root@localhost project_quota_test1]# xfs_quota -c df
Filesystem           1K-blocks       Used  Available  Use% Pathname
/dev/sdb              83845120     157980   83687140    0% /xfs_local
/dev/sdb                102400     124928 9223372036854753280  122% /xfs_local/project_quota_test1
[root@localhost project_quota_test1]# dd if=/dev/zero of=test100MB bs=1M count=100
dd: error writing ‘test100MB’: No space left on device
79+0 records in
78+0 records out
81788928 bytes (82 MB) copied, 0.163849 s, 499 MB/s
[root@localhost project_quota_test1]# xfs_quota -c df
Filesystem           1K-blocks       Used  Available  Use% Pathname
/dev/sdb              83845120     237748   83607372    0% /xfs_local
/dev/sdb                102400     204800 9223372036854673408  200% /xfs_local/project_quota_test1

It's gone negative. That number in hex: 0x7FFFFFFFFFFE7000

What kernel are you using, and can you outline all the way you set
everything up to cause that to occur? Also, what is the output of a
plain 'df -h' when it is in that state?

No that VM is gone so I can’t check it, I’d have to redo the test. Since at the time I was also testing Btrfs stuff I’m going to say it was at the oldest 3.17.1, and could have been 3.18rc1 or rc2.

Chris Murphy
<Prev in Thread] Current Thread [Next in Thread>