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-17177 Lustre <p>DoM read-on-open feature fetch some data from MDT stripe during file open. Meanwhile that is happening always when it sees data on MDT without regards to actual mirror states. That can cause at least useless reads from non-active mirror, at worse stale data read.</p> <p>Considering that the following changes are proposed:</p> <ul> <li>no read-on-open data prefetch for FLR file</li> <li>no DoM lock/size returned if DoM component has stale FLR flag</li> </ul> <p>thus we avoid potential problems and doesn't lose significant benefits, in most cases mirrored file is not having any real performance gain from read-on-open.</p> LU-17177

prevent DoM read-on-open with FLR

Bug Minor Open Unresolved Mikhail Pershin Mikhail Pershin Tue, 10 Oct 2023 09:34:20 +0000 Thu, 19 Oct 2023 09:43:11 +0000 0 3 <p>"Mikhail Pershin &lt;mpershin@whamcloud.com&gt;" uploaded a new patch: <a href="https://review.whamcloud.com/c/fs/lustre-release/+/52613" class="external-link" target="_blank" rel="nofollow noopener">https://review.whamcloud.com/c/fs/lustre-release/+/52613</a><br/> Subject: <a href="https://jira.whamcloud.com/browse/LU-17177" title="prevent DoM read-on-open with FLR" class="issue-link" data-issue-key="LU-17177">LU-17177</a> mdt: prevent read-on-open for FLR file<br/> Project: fs/lustre-release<br/> Branch: master<br/> Current Patch Set: 1<br/> Commit: 350b8904fb3504697944db82e00380c26397caba</p> <p>During patch review Andreas has mentioned:</p> <blockquote><p>My preference is to avoid disabling this feature/optimization to handle a rare use case. I don't think it matters which mirror the pages are attached to, if the source is not stale. <br/> Definitely the DoM data should not be returned if that component is marked stale, but otherwise it should be returned to the client and attached to any mirror.</p></blockquote> <p>That is also possible way to go, but it has to avoid <a href="https://jira.whamcloud.com/browse/LU-16468" title="some miscellaneous IOs need to protect accessing layout" class="issue-link" data-issue-key="LU-16468"><del>LU-16468</del></a> issue which happenned due to missing <tt>lov_io_mirror_init()</tt> I think. So far I see two possibilities:</p> <ul> <li>remove <tt>ci_ignore_layout</tt> in <tt>ll_dom_finish_open()</tt> for <tt>CIT_MISC</tt> IO, thus <tt>lov_io_mirror_init()</tt> will be not skipped in <tt>lov_io_slice_init()</tt></li> <li>keep that flag but do <tt>lov_io_mirror_init()</tt> always, with and without that flag</li> </ul> <p>Initially that <tt>ci_ignore_layout</tt> was used to prevent <tt>layout_refresh()</tt> call while caller is under live layout lock, so it can reveal potential deadlock. That doesn't look so from code, it should match existing lock and proceed but still that flag removal need extra tests, racer probably. On other hand, including <tt>lov_io_mirror_init()</tt> to be called with 'ignore_layout'  flag can be not safe too, when MISC IO is started from OSC</p> Related LU-16468 Development Rank 1|i03xvr: Rank (Obsolete) 9223372036854775807 Severity