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-15433 Lustre <ul> <li>with <a href="https://jira.whamcloud.com/browse/LU-14642" title="transfer layout version to OST objects in layout change" class="issue-link" data-issue-key="LU-14642"><del>LU-14642</del></a> landed MDS doesn't transfer new layout to OST anymore</li> <li>lfs mirror resync doesn't write to primary replica, so primary replica's object still has old layout version</li> <li>evicted from MDS client thinks layout lock is still granted and sends OST_WRITE with old layout version to read primary replica's object which still has old version as well</li> <li>OST modifies that object</li> <li>replicas are out of sync</li> </ul> LU-15433

evicted client can corrupt mirrored file

Bug Minor Open Unresolved WC Triage Alex Zhuravlev flr-improvement Tue, 11 Jan 2022 11:46:39 +0000 Tue, 26 Jul 2022 10:33:38 +0000 0 6 <p>Is this really possible? The MDS would mark the other replicas stale as soon as the client tried to write, so the client would have to be evicted, "<tt>lfs mirror resync</tt>" run and finish on the file to clear the STALE flag from the mirrors, and <b>then</b> the client would write to the primary before it detected that it was evicted? It seems like a long time for the client to not detect that it is evicted?</p> <p>One possible solution would be for the MDS to bump the layout version of a file if it evicts a client that is writing to it. Then the evicted client would have to re-fetch the layout before it could write again. Alternately, flag the primary with an "EVICTED" flag that only triggers the layout version to be increased when "<tt>lfs mirror resync</tt>" is actually run. That would avoid bumping the version repeatedly and hurting other clients writing to the same file, without any danger that the STALE flag is cleared.</p> <p>that's the point AFAIU - it takes time for an evicted client to notice eviction and this window is more than enough for something like resync if the file is small enough?<br/> another important thing is that <a href="https://jira.whamcloud.com/browse/LU-14642" title="transfer layout version to OST objects in layout change" class="issue-link" data-issue-key="LU-14642"><del>LU-14642</del></a> changes the original model where layout version was distributed by MDS to the model where the distribution is done by client. I guess Bobi Jam can comment on this better.</p> <p>Given that most of the fixes for <a href="https://jira.whamcloud.com/browse/LU-14642" title="transfer layout version to OST objects in layout change" class="issue-link" data-issue-key="LU-14642"><del>LU-14642</del></a> have now been merged into maste / master-next what do you guys want to do with this issue??</p> <p>will try <a href="https://review.whamcloud.com/#/c/46707/" class="external-link" target="_blank" rel="nofollow noopener">https://review.whamcloud.com/#/c/46707/</a> again</p> Development Epic Link EX-4394 Rank 1|i02ebz: Rank (Obsolete) 9223372036854775807 Severity