https://jira.whamcloud.com This file is an XML representation of an issue en-us 9.4.14 940014 05-12-2023 https://jira.whamcloud.com/browse/LU-5155 Lustre <p>We would like a way to map object numbers in ZFS back to filenames in the Lustre filesystem.</p> <p>This will be useful in a number of cases, but a common one will be when ZFS detects corrupt files, and we need to figure out which Lustre file is effected.</p> LU-5155

Need a way to map ZFS OSD objects to Lustre filenames

Improvement Major Resolved Fixed Alex Zhuravlev Christopher Morrone zfs Fri, 6 Jun 2014 18:02:27 +0000 Tue, 17 Mar 2015 12:45:27 +0000 Tue, 17 Mar 2015 12:45:27 +0000 Lustre 2.8.0 0 7 <p>Hi Alex</p> <p>Could you please advise on this one?</p> <p>Thanks</p> <p>Peter</p> <p>This should hopefully be pretty easy. Each OST object holds the parent MDT inode FID in the "fid" xattr and then "lfs fid2path /mount $(parent FID)" can be used to generate the pathname on the on any client. </p> <p>On ldiskfs it is possible to access the "fid" xattr on a mounted OST via:</p> <div class="preformatted panel" style="border-width: 1px;"><div class="preformattedContent panelContent"> <pre>debugfs -c -R 'stat /O/$seq/d$((oid % 32))/$oid' /dev/ostdev </pre> </div></div> <p>or on the ldiskfs-mounted OST filesystem via the ll_decode_filter_fid utility.</p> <p>On ZFS it is possible that ll_decode_filter_fid may work, but it isn't possible to mount the OST concurrently between Lustre and ZPL to access the fid xattr on an object.</p> <p>It <em>might</em> be possible to extend the "FID2PATH" ioctl interface to do the OST FID to MDT parent FID mapping directly from the client and have it continue to generate the full path, but I wouldn't want to commit to that without first looking at the code in detail.</p> <p>Another less complex option would be to extend l_decode_filter_fid to do an ioctl locally on the OST to look up the MDT parent FID so that it can be used by fid2path. </p> <p>Thanks, I was able to get the fid from the xattr using zdb. I made a little python script to call zdb and convert the string into standard Lustre FID text form. It is here for reference:</p> <p><a href="https://raw.githubusercontent.com/chaos/lustre-tools-llnl/1.8/scripts/zfsobj2fid" class="external-link" target="_blank" rel="nofollow noopener">https://raw.githubusercontent.com/chaos/lustre-tools-llnl/1.8/scripts/zfsobj2fid</a></p> <p>Isaac Huang (he.huang@intel.com) uploaded a new patch: <a href="http://review.whamcloud.com/13721" class="external-link" target="_blank" rel="nofollow noopener">http://review.whamcloud.com/13721</a><br/> Subject: <a href="https://jira.whamcloud.com/browse/LU-5155" title="Need a way to map ZFS OSD objects to Lustre filenames" class="issue-link" data-issue-key="LU-5155"><del>LU-5155</del></a> scripts: added contrib/scripts/zfsobj2fid.py<br/> Project: fs/lustre-release<br/> Branch: master<br/> Current Patch Set: 1<br/> Commit: e27d0ae3328fbccf5a250bee0ad26dc2c9fc4621</p> <p>A better long-term solution would be to extend the fid2path code to accept an OST object FID directly, have it fetch the trusted.fid xattr from the OST object (possibly at the MDS via OUT RPC as LFSCK does), extract the parent inode FID, and proceed with normal fid2path processing (maybe returning -EREMOTE to the client if the inode FID is on a different MDT and continuing on a different MDT).</p> <p>Oleg Drokin (oleg.drokin@intel.com) merged in patch <a href="http://review.whamcloud.com/13721/" class="external-link" target="_blank" rel="nofollow noopener">http://review.whamcloud.com/13721/</a><br/> Subject: <a href="https://jira.whamcloud.com/browse/LU-5155" title="Need a way to map ZFS OSD objects to Lustre filenames" class="issue-link" data-issue-key="LU-5155"><del>LU-5155</del></a> scripts: added lustre/scripts/zfsobj2fid<br/> Project: fs/lustre-release<br/> Branch: master<br/> Current Patch Set: <br/> Commit: 2ccda307213e93439fb3e182eb2618f8dac21419</p> <p>Landed for 2.8</p> Development Rank 1|hzwnwf: Rank (Obsolete) 14223