John Amodeo schrieb: > > Greetings, > > I have noticed some strange behavior with the way Cyrus interacts with > the Synchronous bit on Ext2 filesystems. This may or may not be a > problem. I am not sure. I am looking for some help or advice. > > Installation directions imply that setting the Synchronous bit on the > following directories: > > cd /var/imap > chattr +S . user quota > chattr +S /var/spool/imap > > makes Cyrus operate more reliably on Ext2, and may even fix locking > issues, etc. etc. So I decided to give it a shot. > > >From what I understand, when you set a Synchronous bit on a parent > directory, any files or sub directories that are created also inherit > the synchronous bit. I tested this and it is true. If I set the +S on > the mailstore directory, go into it, and create a directory called > "test", an "lsattr test" shows the sync bit is set. Furthermore, > creating a file in this directory also shows the same. > > But when you use cyradm to create a new user's mailbox, the +S bit is > not set on the new directory. If you type "mkdir <useraccount>", the +S > bit *is* set. It seems the inheritance only occurs when the files are > created in a Unix shell, not through a daemon such as Imap or something.
Interesting. I don't know too much about the +S bit but I have the same problems with directories where the sticky bit is set on the group to get group inheritance. When I create a subdir with mkdir or via samba, the group is the same like the parent. When I'm using cp to copy an existing dir, the sticky bit is ignored. I know it even depends whether you use cp and tar as root or not. But it's sometimes difficult to understand. -Simon > > Also, new mail that gets delivered to a user's Inbox does not have the > +S, but if you create a new file in the user's Inbox, it will inherit > the +S... > > If this is true, and not actually a bug or something I am doing wrong, > doesn't this throw the whole Synchronous theory out the window? > > Thoughts? > > -John
