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-13078 Lustre <p>It would a great feature for the MGS to ask/force a client to umount. This can be very useful in very large clusters with multiple networks.</p> LU-13078

mgs trigger umount of clients

New Feature Minor Open Unresolved Peter Jones Mahmoud Hanafi Fri, 13 Dec 2019 16:48:47 +0000 Sat, 11 Sep 2021 07:55:40 +0000 0 4 <p>This is similar to a feature which has been discussed in the past under the name of "<a href="http://wiki.lustre.org/Simplified_Interoperability" class="external-link" target="_blank" rel="nofollow noopener">Simplified Interoperability</a>". In that feature the intent is to quiesce clients during a server failover or upgrade. There is already an architecture document for this feature, and surprisingly an old patch in the <a href="https://bugzilla.lustre.org/show_bug.cgi?id=18496" class="external-link" target="_blank" rel="nofollow noopener">bugzilla 18496 ticket</a> as well, though it is against Lustre 2.0 and would obviously need to be updated, but may at least serve as a guide for implementation. </p> <p>One difference in this request vs. Simplified Interoperability is that unmounting a client may be problematic due to open file handles, unlike the original proposal which only quiescent the client mountpoints. The VFS will prevent the unmount with open files due to reference counting on the filesystem mountpoint. In newer kernels it <em>may</em> be possible to abort/evict the file handles from the filesystem side, but this would cause application errors. </p> Duplicate LU-5703 Development Rank 1|i00r0n: Rank (Obsolete) 9223372036854775807