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-16493 Lustre <p>the subtest expects 50 files each with 1K stripes can be created in 60 seconds which is not quite enough for ZFS. I did local measurements and observed:</p> <ul> <li>OST object precreation doesn't block creations (MDS has enough objects at any moment)</li> <li>MDS needs to update last_used file for each OST object used, i.e. 50*1000=50K times</li> <li>ZFS's writerecord and writerecord's declaration is still much slower compared to ldiskfs (2-3x slower)</li> <li></li> </ul> LU-16493

recovery-small/144b fails with ZFS

Bug Trivial Resolved Fixed Alex Zhuravlev Alex Zhuravlev Thu, 19 Jan 2023 14:10:57 +0000 Tue, 14 Feb 2023 13:38:38 +0000 Tue, 14 Feb 2023 13:38:38 +0000 0 2 <p>"Alex Zhuravlev &lt;bzzz@whamcloud.com&gt;" uploaded a new patch: <a href="https://review.whamcloud.com/c/fs/lustre-release/+/49693" class="external-link" target="_blank" rel="nofollow noopener">https://review.whamcloud.com/c/fs/lustre-release/+/49693</a><br/> Subject: <a href="https://jira.whamcloud.com/browse/LU-16493" title="recovery-small/144b fails with ZFS" class="issue-link" data-issue-key="LU-16493"><del>LU-16493</del></a> tests: recovery-small/144b to wait longer<br/> Project: fs/lustre-release<br/> Branch: master<br/> Current Patch Set: 1<br/> Commit: ead9f854705427793091837945e7576ff79b96a2</p> <p>did few measurements with <a href="https://jira.whamcloud.com/browse/LU-13139" title="collect stats for each OSD method" class="issue-link" data-issue-key="LU-13139">LU-13139</a>:</p> <div class="code panel" style="border-width: 1px;"><div class="codeContent panelContent"> <pre class="code-java"> ldiskfs zfs-op zfs-noop osd decl write 1 2 274 osd write 44 55 127 lod declare create 108577. 139043 396156 lod create 57952. 90884 171936 </pre> </div></div> <p>cost of DTAPI calls in usec, where zfs-op is master + <a href="https://jira.whamcloud.com/browse/LU-13132" title="osd-zfs to cache dbufs" class="issue-link" data-issue-key="LU-13132"><del>LU-13132</del></a> + <a href="https://jira.whamcloud.com/browse/LU-14918" title="too many ldiskfs transaction credits when unlinking overstriped files" class="issue-link" data-issue-key="LU-14918"><del>LU-14918</del></a>, zfs-noop is clean master.<br/> for each stripe OSP will be calling osd_declare_write() and osd_write(), this leads to much more expensive lod's declaration and creation.</p> <p>"Oleg Drokin &lt;green@whamcloud.com&gt;" merged in patch <a href="https://review.whamcloud.com/c/fs/lustre-release/+/49693/" class="external-link" target="_blank" rel="nofollow noopener">https://review.whamcloud.com/c/fs/lustre-release/+/49693/</a><br/> Subject: <a href="https://jira.whamcloud.com/browse/LU-16493" title="recovery-small/144b fails with ZFS" class="issue-link" data-issue-key="LU-16493"><del>LU-16493</del></a> tests: recovery-small/144b to wait longer<br/> Project: fs/lustre-release<br/> Branch: master<br/> Current Patch Set: <br/> Commit: 1c8b40d5e445fac4b4809e082f622efb642e3da5</p> <p>Landed for 2.16</p> Development Rank 1|i03amf: Rank (Obsolete) 9223372036854775807 Severity