Stefan.O.Nordlander@AstraZeneca.com
2007-Jan-18 21:16 UTC
[Lustre-devel] [Bug 11563] unknown fcntl lock type: 96 when writing files using samba(?)
Please don''t reply to lustre-devel. Instead, comment in Bugzilla by using the following link: https://bugzilla.lustre.org/show_bug.cgi?id=11563 Created an attachment (id=9368) Please don''t reply to lustre-devel. Instead, comment in Bugzilla by using the following link: --> (https://bugzilla.lustre.org/attachment.cgi?id=9368&action=view) unknown fcntl lock dump
martin.budsjo@astrazeneca.com
2007-Jan-19 02:50 UTC
[Lustre-devel] [Bug 11563] unknown fcntl lock type: 96 when writing files using samba(?)
Please don''t reply to lustre-devel. Instead, comment in Bugzilla by using the following link: https://bugzilla.lustre.org/show_bug.cgi?id=11563 OK. We had to upgrade to 1.4.8 due to the problem with zero padded files as reported in bug #11527. But 1.4.8 changes the behaviour for some fctl calls. Now all flock calls returns ENOSYS by default. This breaks one of our applications. This application works just fine with the filesystem remounted with the -o flock option, but that breaks WingIde and Samba. (And perhaps a couple of other applications...) We need a workaround fast. Could we downgrade the clients to 1.4.7 to avoid the problem? Or could the bug #11527 be backed out in some way? (It forces usage of the flock mount option when it''s not ready) The attached log file is from a SMB (samba) process crash. The process itself gets stuck in D state (un-interruptable sleep) Regards Martin Budsjö
martin.budsjo@astrazeneca.com
2007-Jan-19 04:10 UTC
[Lustre-devel] [Bug 11563] unknown fcntl lock type: 96 when writing files using samba(?)
Please don''t reply to lustre-devel. Instead, comment in Bugzilla by using the following link: https://bugzilla.lustre.org/show_bug.cgi?id=11563 The bug to be backed out is bug #10743 not bug #11572 as stated above.
adilger@clusterfs.com
2007-Jan-19 12:34 UTC
[Lustre-devel] [Bug 11563] unknown fcntl lock type: 96 when writing files using samba(?)
Please don''t reply to lustre-devel. Instead, comment in Bugzilla by using the following link: https://bugzilla.lustre.org/show_bug.cgi?id=11563 In reply to comment #3)> OK. We had to upgrade to 1.4.8 due to the problem with zero padded files as > reported in bug #11527. But 1.4.8 changes the behaviour for some fctl calls. > Now all flock calls returns ENOSYS by default. This breaks one of our > applications. This application works just fine with the filesystem remounted > with the -o flock option, but that breaks WingIde and Samba. (And perhaps a > couple of other applications...) > > We need a workaround fast. Could we downgrade the clients to 1.4.7 to avoid > the problem? Or could the bug #11527 be backed out in some way? (It forces > usage of the flock mount option when it''s not ready)It should be possible to downgrade the clients to 1.4.7.3 to avoid the flock issues - the 10730 bug only affects servers.
martin.budsjo@astrazeneca.com
2007-Jan-20 15:50 UTC
[Lustre-devel] [Bug 11563] unknown fcntl lock type: 96 when writing files using samba(?)
Please don''t reply to lustre-devel. Instead, comment in Bugzilla by using the following link: https://bugzilla.lustre.org/show_bug.cgi?id=11563 (In reply to comment #3)> In reply to comment #3) > > We need a workaround fast. Could we downgrade the clients to 1.4.7 to avoid > > the problem? Or could the bug #11527 be backed out in some way? (It forces > > usage of the flock mount option when it''s not ready) > > It should be possible to downgrade the clients to 1.4.7.3 to avoid the flock > issues - the 10730 bug only affects servers.Good.. We will try that. What about the future releases? Does 1.6 behave as 1.4.8 regarding flock support?