[LU-1842] Quota enforcement landing

• Type: New Feature
• Resolution: Fixed
• Priority: Minor
• Fix Version/s: Lustre 2.4.0, Lustre 2.1.4, Lustre 1.8.9
• Affects Version/s: Lustre 2.4.0
• Assignee: Johann Lombardi (Inactive)
• Reporter: Johann Lombardi (Inactive)

Description

Tracker bug for landing orion_quota to master.

Issue Links

• is duplicated by: LU-1300 Quota enforcement support for OSD Restructuring
• is related to: LU-5975 Consider removing quotacheck compatibility code
• is related to: LU-5978 Remove migration of old admin quota files
• is related to: LU-5980 Eliminate old quota formats
• is related to: LUDOC-89 Quota Enforcement Doc Changes
• is related to: LU-1789 Lightweight connection support
• is related to: LU-1318 Quota Enforcement DLM&FID

Comments

[chas williams - CONTRACTOR]

how clever. apparently i have to have the quota_v2 module loaded for this feature. however, ldiskfs doesnt depend on this module and doesnt seem to automagically load it either. this doesnt appear to be happening because QFMT_VFS_V1 isn't mapped to any particular module name in linux/quota.h:

#define INIT_QUOTA_MODULE_NAMES {\

{QFMT_VFS_OLD, "quota_v1"}

,\

{QFMT_VFS_V0, "quota_v2"}

,\
{0, NULL}}

[Johann Lombardi (Inactive)]

Quota enforcement actually requires accounting to be enabled. The only possible combinations are:

• no accounting & no limit enforcement
• space accounting enabled & no limit enforcement
• space accounting enabled & quota limit enforced

[chas williams - CONTRACTOR]

i ported the ldiskfs patches to sles11sp2's kernel (basically 3.0). so it is possible to run quota and usage accounting at the same time? i guess i need to look harder at the ldiskfs patches.

[Johann Lombardi (Inactive)]

This code snippet just enables space accounting automatically at mount time if the quota feature is enabled on the backend ldiskfs filesystem.

Could you please tell us more about the problem you see with ldiskfs and quota in 2.4?

[chas williams - CONTRACTOR]

Hello,

I am seeing a problem with ldiskfs and quotas in 2.4. http://review.whamcloud.com/4098 turned on quotas by default when mounting. However, when my ldiskfs goes to mount, it will fails since quotas and usage accounting are contraindicated. from the "first class accounting" ldiskfs patch:

+#ifdef CONFIG_QUOTA
+ /* Enable space tracking during mount, enforcement can be enabled/disable
+ * later with quota_on/off */
+ if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
+ !(sb->s_flags & MS_RDONLY)) {
+ ret = ext4_acct_on(sb);
+ if (ret)

{ + ext4_msg(sb, KERN_ERR, "Can't enable usage tracking on " + "a filesystem with the QUOTA feature set"); + goto failed_mount4; + }

+ }

am i confused about what is going on here?

[James A Simmons]

Appears master + http://review.whamcloud.com/#change,4218 fixed my problem.

[Johann Lombardi (Inactive)]

For the record, LUDOC-89 is the bug to track changes to the manual.

[Johann Lombardi (Inactive)]

That said, there is still a bug related to oi scrub since it should definitely properly clean up if the quota slave instance failed to be started. I will talk to Fanyong.

Patch for the oi scrub init issue (independent of quota) is available here:
b2_3: http://review.whamcloud.com/#change,4218
master: http://review.whamcloud.com/#change,4219

[Niu Yawei (Inactive)]

utils: tunefs.lustre --quota
http://review.whamcloud.com/420

[Johann Lombardi (Inactive)]

For the record, i have verified that the patch fixes the problem.
Thanks James!

[Johann Lombardi (Inactive)]

Hi James,

Could you please give a try to this patch? http://review.whamcloud.com/4203

It should address the qsd setup issue.
That said, there is still a bug related to oi scrub since it should definitely properly clean up if the quota slave instance failed to be started. I will talk to Fanyong.

Thanks in advance.