Bug 1309498
Summary: | Rebase xfsprogs; enable metadata CRCs by default | ||
---|---|---|---|
Product: | Red Hat Enterprise Linux 7 | Reporter: | Eric Sandeen <esandeen> |
Component: | xfsprogs | Assignee: | Eric Sandeen <esandeen> |
Status: | CLOSED ERRATA | QA Contact: | Zorro Lang <zlang> |
Severity: | unspecified | Docs Contact: | Milan Navratil <mnavrati> |
Priority: | medium | ||
Version: | 7.3 | CC: | dchinner, dshaw, dwysocha, eguan, esandeen, galileo, martin, mnavrati, swhiteho |
Target Milestone: | rc | Keywords: | FutureFeature, Rebase |
Target Release: | --- | ||
Hardware: | Unspecified | ||
OS: | Unspecified | ||
Whiteboard: | |||
Fixed In Version: | xfsprogs-4.5.0-1.el7 | Doc Type: | Release Note |
Doc Text: |
_xfsprogs_ rebased to version 4.5.0
The _xfsprogs_ packages have been upgraded to upstream version 4.5.0, which provides a number of bug fixes and enhancements over the previous version. The Red Hat Enterprise Linux 7.3 kernel RPM requires the upgraded version of _xfsprogs_ because the new default on-disk format requires special handling of log cycle numbers when running the *xfs_repair* utility. Notable changes include:
* Metadata cyclic redundancy checks (CRCs) and directory entry file types are now enabled by default. To replicate the older *mkfs* on-disk format used in earlier versions of Red Hat Enterprise Linux 7, use the "-m crc=0 -n ftype=0" options on the *mkfs.xfs* command line.
* The `GETNEXTQUOTA` interface is now implemented in *xfs_quota*, which allows fast iteration over all on-disk quotas even when the number of entries in the user database is extremely large.
Also, note the following differences between upstream and Red Hat Enterprise Linux 7.3:
* The experimental sparse inode feature is not available.
* The free inode btree (finobt) feature is disabled by default to ensure compatibility with earlier Red Hat Enterprise Linux 7 kernel versions.
|
Story Points: | --- |
Clone Of: | Environment: | ||
Last Closed: | 2016-11-04 06:24:00 UTC | Type: | Bug |
Regression: | --- | Mount Type: | --- |
Documentation: | --- | CRM: | |
Verified Versions: | Category: | --- | |
oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |
Cloudforms Team: | --- | Target Upstream Version: | |
Embargoed: | |||
Bug Depends On: | |||
Bug Blocks: | 1306029, 1340553 |
Description
Eric Sandeen
2016-02-18 00:13:47 UTC
*** Bug 1309501 has been marked as a duplicate of this bug. *** https://access.redhat.com/solutions/2120061 This response states: "RHEL7 is using XFS v5 because xfsprogs-3.2.x uses it as default. xfs filesystem created with -m crc=1 is a production ready feature from RHEL 7.1 and onwards." Unfortunately that is not completely correct, by default mkfs.xfs will create a V4 xfs file system. It would be great to get a V5 xfs during install also. Thanks for the heads up, I've added a comment to that solution, I'll see if I can get it fixed. (In reply to Eric Sandeen from comment #0) > We've been shipping metadata-crc capable xfsprogs since RHEL7 GA, but have > not yet turned them on. There are good reasons to do so now, and they have > been default upstream for quite some time. Let's rebase xfsprogs to > upstream, and gain this capability. NOTE: I do not know much about this feature yet. However, my initial thoughts are as follows. It would be good to know what the "good reasons" are to change the default now. I'm nervous about it and not sure customers have requested it specifically. However, if we know of data corruption cases which would have been avoided should this been enabled we can say they implicitly might want this. I'm sure there's data safety upsides but likely at least some sensitive customers will see performance hit. There's also the possible further complication with the new 'finobt'. This can all be handled and I'm not saying we should not do this but I'm voicing my concerns here from support perspective. Ok, this could be a long list. We went from 3.2.2 to 4.5.0, so I'll stick to the highlights. Notable things from the upstream changelogs: Metadata CRCs, free inode btrees, and directory entry filetypes are now enabled by default. To replicate the older mkfs on-disk format used in prior RHEL7 releases, use the "-m crc=0 -n ftype=0" options on the mkfs.xfs commandline. The GETNEXTQUOTA interface is implemented in xfs_quota. This allows fast iteration over all on-disk quotas even when the number of entries in the user database is extremely large. The RHEL7.3 kernel rpm requires this version of xfsprogs, because the new default on-disk format requires special handling of log cycle numbers during xfs_repair, which is provided in this release. (In reply to Eric Sandeen from comment #10) > Ok, this could be a long list. We went from 3.2.2 to 4.5.0, so I'll stick > to the highlights. > > Notable things from the upstream changelogs: > > Metadata CRCs, free inode btrees, and directory entry filetypes are now > enabled by default. IIRC, the free inode btree feature is not enabled by default in latest RHEL7.3 xfsprogs [root@dhcp-66-87-213 sanity]# uname -a Linux dhcp-66-87-213.rhts.eng.nay.redhat.com 3.10.0-470.el7.x86_64 #1 SMP Fri Jul 15 18:32:03 EDT 2016 x86_64 x86_64 x86_64 GNU/Linux [root@dhcp-66-87-213 sanity]# rpm -q xfsprogs xfsprogs-4.5.0-5.el7.x86_64 [root@dhcp-66-87-213 sanity]# mkfs.xfs -f /dev/vda6 meta-data=/dev/vda6 isize=512 agcount=4, agsize=589824 blks = sectsz=512 attr=2, projid32bit=1 = crc=1 finobt=0, sparse=0 data = bsize=4096 blocks=2359296, imaxpct=25 = sunit=0 swidth=0 blks naming =version 2 bsize=4096 ascii-ci=0 ftype=1 log =internal log bsize=4096 blocks=2560, version=2 = sectsz=512 sunit=0 blks, lazy-count=1 realtime =none extsz=4096 blocks=0, rtextents=0 Yes, finobt=0 by default :) Fixed - thanks Eryu! -Eric XFS crc is enabled as default on xfsprogs-4.5.0-6.el7: # mkfs.xfs -d name=file.img,file=1,size=1g meta-data=file.img isize=512 agcount=4, agsize=65536 blks = sectsz=512 attr=2, projid32bit=1 = crc=1 finobt=0, sparse=0 data = bsize=4096 blocks=262144, imaxpct=25 = sunit=0 swidth=0 blks naming =version 2 bsize=4096 ascii-ci=0 ftype=1 log =internal log bsize=4096 blocks=2560, version=2 = sectsz=512 sunit=0 blks, lazy-count=1 realtime =none extsz=4096 blocks=0, rtextents=0 And crc isn't enabled as default on xfsprogs-3.2.2-2.el7 # mkfs.xfs -d name=file.img,file=1,size=1g meta-data=file.img isize=256 agcount=4, agsize=65536 blks = sectsz=512 attr=2, projid32bit=1 = crc=0 finobt=0 data = bsize=4096 blocks=262144, imaxpct=25 = sunit=0 swidth=0 blks naming =version 2 bsize=4096 ascii-ci=0 ftype=0 log =internal log bsize=4096 blocks=2560, version=2 = sectsz=512 sunit=0 blks, lazy-count=1 realtime =none extsz=4096 blocks=0, rtextents=0 So verified this bug. And due to this bug do rebase from xfsprogs-3.2 to xfsprogs-4.5, so I've done XFS regression test to make sure there's no regression bug bring in by this change. (I think we don't need to write a test case to cover this bug.) Thanks, Zorro Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory, and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. https://rhn.redhat.com/errata/RHBA-2016-2431.html |