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-2016 Lustre <p>Add the possibility to swap the layouts between to files.<br/> This feature will be used to migrate files inside Lustre (eg for rebalance) and by HSM to restore files from archive<br/> This patch is the mdd part of the feature, the high level design is:<br/> – Start transaction<br/> – Read LOV EA from obj1 &amp; obj 2<br/> – Run xattr_set(obj1, layout2) &amp; xattr_set(obj2, layout1)<br/> (LOD should not take any action)<br/> – Stop transaction</p> LU-2016

Layout swapping, MDD part

New Feature Minor Resolved Fixed Jodi Levi CEA HSM Mon, 24 Sep 2012 04:49:23 +0000 Wed, 5 Aug 2020 13:32:55 +0000 Sat, 13 Apr 2013 15:44:22 +0000 Lustre 2.4.0 Lustre 2.4.0 0 7 <p>Jodi will see that this ticket is appropriately assigned</p> <p>Patch is at <a href="http://review.whamcloud.com/4189" class="external-link" target="_blank" rel="nofollow noopener">http://review.whamcloud.com/4189</a></p> <p>I was thinking that a good sanityt test for the layout swap and layout lock code would be to implement an "lfs swap" command and hook this into the "lfs_migrate" script instead of using "mv". The core "lfs swap"/"lfs_swap" command functionality should be implemented via new llapi helpers:</p> <div class="code panel" style="border-width: 1px;"><div class="codeContent panelContent"> <pre class="code-java"><span class="code-object">int</span> llapi_fd_layout_swap(<span class="code-object">int</span> src_fd, <span class="code-object">int</span> tgt_fd, __u64 flags) { struct ll_layout_swap lls = { .lls_tgt = tgt_fd, .lls_flags = flags }; <span class="code-keyword">return</span> ioctl(fd, LL_IOC_LOV_LAYOUT_SWAP, (void *)&amp;lls) } <span class="code-object">int</span> llapi_file_layout_swap(<span class="code-keyword">const</span> <span class="code-object">char</span> *src_path, <span class="code-keyword">const</span> <span class="code-object">char</span> *tgt_path, __u64 flags) { <span class="code-object">int</span> src_fd, tgt_fd; src_fd = open(src_path, O_RDONLY); tgt_fd = open(tgt_path, O_RDONLY); <span class="code-comment">/* error handling and such */</span> <span class="code-keyword">return</span> llapi_fd_layout_swap(src_fd, tgt_fd, flags); } </pre> </div></div> <p>Where ll_file_ioctl(LL_IOC_LOV_LAYOUT_SWAP) triggers an MDS RPC to call mdd_swap_layouts().</p> <p>"lfs swap" wouldn't be 100% atomic yet, since the copy + swap would not yet be safe from other thread modifying the original object during the copy, but it would at least fix the problem of preserving open file handles on the original file and would be a good step toward implementing a full atomic "lfs migrate".</p> <p>A test script for this code could look like:</p> <ol> <li>generate 10s of verifiable IO data<br/> dd if=/dev/urandom of=$DIR/$tfile.orig bs=1M &amp;<br/> DD_PID=$?<br/> sleep 10<br/> kill $DD_PID</li> </ol> <ol> <li>copy the data, while swapping the layout repeatedly and</li> <li>changing the stripe count to exercise the IO/LOV code<br/> dd if=$DIR/$tfile.orig of=$DIR/$tfile.copy bs=1M &amp;<br/> DD_PID=$?<br/> sleep 0.5<br/> while kill -STOP $DD_PID; do<br/> lfs_migrate -y -c $((RANDOM % OSTCOUNT)) $DIR/$tfile.copy<br/> kill -CONT $DD_PID<br/> sleep 1<br/> done</li> </ol> <ol> <li>verify that the migrated file matches the original</li> <li>lfs_migrate itself verifies each copy after it is made<br/> cmp $DIR/$tfile.orig $DIR/$tfile.copy || error "compare failed"</li> </ol> <p>At some later point (or immediately, if someone is ambitious) lfs_migrate would be implemented with "lfs migrate", which calls lfs_setstripe() internally to process the arguments to create the target object(s), then gets the group lock and does the data copy internally before calling llapi_fd_layout_swap() to do the swap (like an HSM copytool does). At this point, the above "kill -STOP" can be replaced with "kill -0" (just to check that "dd" is still running) and should be able to migrate the file while it is being written.</p> <p>I am implementing lfs swap_layouts in <a href="https://jira.whamcloud.com/browse/LU-2017" title="Layout swapping, client and MDT parts" class="issue-link" data-issue-key="LU-2017"><del>LU-2017</del></a>, I should push a first version next week.<br/> I have made a simple test without concurrent access, I will add the concurrent access test.</p> <p>About the open, today I open with O_WRONLY in place of O_RDONLY because swaping is changing file content.<br/> With O_RDONLY someone can swap file it should not be able to modify.</p> <p>Change 4189 is landed, can this be closed?</p> Related LU-2017 LU-3608 Development Project Rank 1|hzv3of: Rank (Obsolete) 4118