Hello, currently we are storing the incoming mail in the user's home directory, which is within a distributed filesystem called AFS (OpenAFS). Accessing the "mbox" file directly (using the distributed filesystem) works fine with mutt, thunderbird, ... To access the INBOX from a machine, which does not have access to this distributed filesystem, we are using IMAP (UW-imap). Now I wanted to check if dovecot can do the job (maybe faster than UW), and I had the following problem: This is my configuration:
# 1.1.4: /opt/dovecot-1.1.4/etc/dovecot.conf log_path: /var/log/dovecot ssl_cert_file: /etc/ssl/certs/imapd.pem ssl_key_file: /etc/ssl/keys/imapd.pem login_dir: /var/run/dovecot/login login_executable: /opt/dovecot-1.1.4/libexec/dovecot/imap-login first_valid_uid: 100 mail_location: mbox:~/Mail:INBOX=~/.maildir/incoming mail_debug: yes mmap_disable: yes mail_nfs_storage: yes mail_nfs_index: yes mbox_write_locks: fcntl mbox_dirty_syncs: no mbox_lazy_writes: no mail_executable: /opt/dovecot-1.1.4/libexec/dovecot/afs-token /opt/dovecot-1.1.4/libexec/dovecot/imap auth default: verbose: yes debug: yes passdb: driver: pam args: setcred=yes session=yes dovecot userdb: driver: passwd
When I start my imap-client, I see the content of the INBOX, and I can read, delete, ... the different entries. But when new mail arrives dovecot (and my imap-client) will never notice this. And when I quit my imap-client the dovecot imap server will write back its view of the INBOX, and delete this new mail. This seems to be a problem related to AFS. I poked in the dovecot code and have the following theory: dovecot opens the INBOX R/W, dovecot calls "stat" on the INBOX file periodically to look for modifications. But this does not work, because this machine will stat the local copy (AFS cache) of the INBOX as long as this file is still open R/W. My questions: Is it possible to have a configuration, that the INBOX file is not left open when stat-ing this file? Or is it possible to open the INBOX file R/W only when it us locked? Or is it necessary to modify the code?
Thank you for any help, HW
-- Hans-Werner Paulsen hans@MPA-Garching.MPG.DE MPI für Astrophysik Tel 089-30000-2602 Karl-Schwarzschild-Str. 1 Fax 089-30000-2235 D-85741 Garching