- dovecot - dovecot.org

[Dovecot] Wanting to seperate MTA and Imap
by Brent Clark 26 Nov '07

26 Nov '07

Hi Current my set up is that the MTA and Dovecot (maildir) is on the same machine. Ive been asked to seperate the services from the same machine (too would like to add a secondary MTA - but thats future ideas). I read the documentation on NFS, but from other sites i have read. Most say this is NOT the best solution. My question is, how do some of the fellow dovecot users,or other companys, e.g. google etc, pull this off. If you can assist in helping me fine a solution and too help me understand, I would greatfully appreciate it. Kind Regards Brent Clark

5 4

[Dovecot] libwrap-ing IMAP and POP logins
by Edgar Fuß 26 Nov '07

26 Nov '07

Since I've been using this for maybe a year now, maybe someone else is interested in restricting IMAP and POP logins via libwrap. In addition to the attached patch (against 1.0.5) to src/login-common/main.c, src/{imap,pop3}-login/Makefile.in have to be modified to link against libwrap. Of course, the option needs to be integrated into configure in the long run.

2 2

[Dovecot] Sieve regex never matches
by frank.kintrup＠fkware.de 26 Nov '07

26 Nov '07

Hi, I found the following code on the net which should redirect all mails to another address depending on the time of day (useful for forwarding mails to work): require ["regex"]; if allof (header :regex "date" "(08|09|10|11|12|13|14|15|16|17):..:..", not header :regex "date" "(sat|sun)") { redirect "address@work>"; keep; } To check and test I simplified the regex to match all mails within a certain hour: require ["regex"]; if header :regex "date" "20:..:.." { redirect "user@domain"; keep; } Both scripts compile fine, but their regex never match. Has anyone an idea why? I'm using Dovecot 1.0.0 on Debian Etch stable. Frank

3 2

[Dovecot] passdb ldap static values
by Harrison Metzger 26 Nov '07

26 Nov '07

Hello, In the pass_attrs field one can map ldap attributes to dovecot attributes. I'm wondering if I can map both ldap attribuets and static attributes to a dovecot value in the same passdb section. Also, I can't seem to find what other values (if any) can belong in the: passdb ldap { args = /path/filename } stanza. Are there any other options I can put in that stanza, I can't seem to find any on the wiki. Harrison Metzger

2 1

[Dovecot] Disconnected: Mailbox is in inconsistent state, please relogin
by Kleyton M. 26 Nov '07

26 Nov '07

Bom dia, Pessoal eu tenho um servidor HP Proliant ML 110 com 4gb de memoria e processador Xeon. Minha distribuição é um Fedora Core 7 com Postfix + Procmail e este chama Spanassasim e Clamav Para IMAP, POP3 utilizo dovecot. Estava tendo o erro de Corrupted dovecot.index, atualizei da versao 1.05 para a 1.07 Agora o mesmo problema ocorre algumas messagens ficam presas na caixa e o usuario nao recebe atraves do outlook 2000 (pop3) e quando logo via IMAP alguns usuarios tem menssagens nao lidas que nao foram entregues por pop3 e no Maillog aparece: Nov 23 09:00:43 golf dovecot: IMAP(everli): Maildir /home/everli/Maildir sync: UID < next_uid (89 < 90, file = msg.BAwd:2,) Nov 23 09:00:43 golf dovecot: IMAP(everli): Disconnected: Mailbox is in inconsistent state, please relogin. Apos isso relogando ou clicando novamente na caixa de entrada e deslogando o dovecot entrega via POP3 essas menssagens nao lidas. Porque isso ? como resolvo esse problema pois a maioria utiliza outlook e não o IMAP Cordialmente; Kleyton Maia dos Santos ____________________________________________________________________________________ Get easy, one-click access to your favorites. Make Yahoo! your homepage. http://www.yahoo.com/r/hs

2 1

[Dovecot] imap process killed with signal 11
by Daniel 26 Nov '07

26 Nov '07

Hi! Using the latest cvs from HEAD, I'm experiencing problems after sending a message and the client is trying upload the message to the appropriate (Sent/Queue) folder. I get only one error in dovecot.log: dovecot: Nov 25 14:21:31 Error: child 19554 (imap) killed with signal 11 So my client disconnects, and asks for the password again, and in dovecot.info I get nothing, but the logs of the new connection attempt. What else info should I provide? System is OpenBSD 4.2 $ dovecot -n # 1.1.beta8: /etc/dovecot.conf base_dir: /var/run/dovecot/ log_path: /var/log/mail/dovecot.log info_log_path: /var/log/mail/dovecot.info protocols: pop3 pop3s imap listen: 212.92.15.17 ssl_cert_file: /etc/ssl/dovecot_cert.pem ssl_key_file: /etc/ssl/private/dovecot_key.pem login_dir: /var/run/dovecot/login login_executable(default): /usr/local/libexec/dovecot/imap-login login_executable(imap): /usr/local/libexec/dovecot/imap-login login_executable(pop3): /usr/local/libexec/dovecot/pop3-login login_user: _dovecot login_processes_count: 1 login_max_processes_count: 10 max_mail_processes: 100 first_valid_uid: 8000 first_valid_gid: 8000 mail_location: maildir:~/Maildir/ mail_debug: yes dotlock_use_excl: yes mail_executable(default): /usr/local/libexec/dovecot/imap mail_executable(imap): /usr/local/libexec/dovecot/imap mail_executable(pop3): /usr/local/libexec/dovecot/pop3 mail_plugins: quota mail_plugin_dir(default): /usr/local/lib/dovecot/imap mail_plugin_dir(imap): /usr/local/lib/dovecot/imap mail_plugin_dir(pop3): /usr/local/lib/dovecot/pop3 imap_client_workarounds(default): tb-extra-mailbox-sep imap_client_workarounds(imap): tb-extra-mailbox-sep imap_client_workarounds(pop3): pop3_client_workarounds(default): pop3_client_workarounds(imap): pop3_client_workarounds(pop3): outlook-no-nuls oe-ns-eoh auth default: mechanisms: plain digest-md5 ntlm user: _dv_auth master_user_separator: * verbose: yes passdb: driver: passwd-file args: /etc/dovecot.passdb_master pass: yes master: yes passdb: driver: passwd-file args: /etc/dovecot.passdb userdb: driver: passwd-file args: /etc/dovecot.userdb socket: type: listen client: path: /var/spool/postfix/private/auth mode: 432 user: _postfix group: _postfix master: path: /var/run/dovecot/auth-master mode: 432 user: vmail group: vmail plugin: quota: maildir quota_rule: *:storage=100M quota_rule2: Trash:storage=50M -- Lévai Dániel Key ID: 0x4AC0A4B1 (wwwkeys.eu.pgp.net) Fingerprint: D037 03B9 C12D D338 4412 2D83 1373 917A 4AC0 A4B1

2 1

[Dovecot] recalculation of maildir Quota
by Ingo Rogalsky 26 Nov '07

26 Nov '07

Hello, I intend to use maildir Quota for measurement only, not for applying restrictions. Therefore I defined rather high values. That's working as expected. Nevertheless it's not useful, because the recalculation time is to high. This takes some minutes for my own mailaccount with 250.000 mails on a NFS storage. This would be to bad, but whilst this time, the Mailclient hangs. Probably our users woulnd't accept this. Can this be done in the background? Furthermore I noticed that recalculation occurs after the file maildirsize gets longer than 30 lines (around 250-300 bytes) and not after 5120 bytes as the wiki states. Do I do/understand something wrong or is this a bug? Probably a nightly cronjob, which does calculation of Maildir-size would be a more useful alternative. Regards, Ingo Rogalsky

5 6

[Dovecot] migrating from mbox to maildir
by Erick Perez 26 Nov '07

26 Nov '07

Hi all, I have a locking problem in a Centos 4.4 linux machine with the following config: dovecot-0.99.11-8.EL4 postfix-2.2.10-1.1.el4 I use virtual mailboxes with postfix currently I have dovecot with mbox as this: default_mail_env = mbox:/var/mail/virtualmail/%d/%n The problem is that mail is being delivered to mailboxes very very slowly (sometimes up to 4 hours). In my logs, I have tons of these messages: mail postfix/virtual[24965]: CFB8878D04F: to=<REMOVED@REMOVED>, relay=virtual, delay=51, status=deferred (mailbox /var/spool/mail/virtualmail/REMOVEDDOMAINNAME/npitti: unable to lock for exclusive access: Resource temporarily unavailable) then someone told me that I need to change dovecot to maildir format. Is there a reference to change format? Or some comments about this? Thanks in advance.

5 6

[Dovecot] overquota message
by Emiliano Gabrielli (aka AlberT) 26 Nov '07

26 Nov '07

Hi, is it possible to customize the overquota message ? I'd like very much if it was possible to use the %d in the postmaster mail address (and overquota message) too ... in virtual domain environment it is very likely .. thanks in advance -- <?php echo ' Emiliano Gabrielli (aka AlberT) ',"\n", ' GrUSP founder - ZCE ',"\n", ' AlberT_at_SuperAlberT_it - www.SuperAlberT.it ',"\n", ' IRC: #php,#AES azzurra.com ',"\n",'ICQ: 158591185'; ?>

2 1

[Dovecot] Squat full text search indexer redesign
by Timo Sirainen 26 Nov '07

26 Nov '07

I was writing this to beta9 announcement, but it grew so huge that I guess a separate mail is better. :) I've spent quite a lot of time this last week redesigning/rewriting Squat indexer. I think the redesign is good enough now that I'll just replace the existing squat with this new one in next release. I don't think many people are using the old squat, and there are also some bugs in it that cause some searches not to find anything (haven't tried to fix it). The old Squat indexes about 3MB/s on my machine while the new one indexes 8MB/s. Or it's actually 10MB/s, but at the end of indexing it does a compression step that merges all fragmented UID lists. That drops the index size something like 5-10%, but it's pretty slow because it has to go through all the fragmented lists. It could probably be optimized to be a lot faster, but I think I'll forget about it for now. Besides being faster there are other improvements in it: - Old squat could answer only 4 character long substring searches. New squat can answer 1..n characters long substring searches, where n can be configurable. The larger the n, the larger the index size. For example 4..10 index sizes are about 35% -> 46% -> 53% -> 59% -> 63% -> 67% -> 70% of the mailbox size (with 32MB of Dovecot mailing list archives). With both old and new squat if the search keyword is longer than n, Dovecot looks up the list of all possible n char long combinations of the word, gets a list of their common UIDs and then reads those mails to verify that the word is really found from them. - New squat can also be configured to index longer words for non-substring searches. This allows it to give a definite list of UIDs where the word is found, so Dovecot doesn't have to open those mails to verify it. However it can't give a definite list of UIDs where the substring isn't found. So it still has to do the n-char-combination lookup described above, and for those returned UIDs which aren't also in the definite-UIDs list Dovecot still has to read them to check if the substring is found or not. This could be skipped for non-standard X-FAST-TEXT and X-FAST-BODY searches which don't try to search substrings. Adding 255 char long non-substring searches to 4 char long substring searches grows the index about 7% (35%->42% for 32MB, 28%->36% for 200MB, 27%->33% for 1,4GB). Unfortunately it also grows memory usage pretty much. For a 200MB mailbox heap usage grows from 34MB to 163MB. For 1,4GB lkml mailbox heap usage grows from 80MB to 700MB. So I'm not sure if this should be enabled by default until something can be done about the memory usage. - The current list of indexed characters are: A-Z, 0-9, @.-+#$%_& and all 8bit chars. When trying to search for non-indexed characters Dovecot has to split the search word. For example "a,b" would search "a" and "b", combine the results and then open those mails to see where it really exists. The same thing happens with "a b". If the search word doesn't contain any indexable characters, Dovecot has to read all mails. This indexable character list could be changed runtime, so if a user keeps searching for some non-indexed character, it could be added to indexed characters and the index be rebuilt. - To be able to give definite replies for BODY searches squat has to keep track of whether a word was found from body or header. This also optimizes HEADER searches because then Dovecot has to search only those mails that have the word in their headers, instead of the mails that have it in either header or body. Unfortunately this makes the indexes somewhat larger again (22%->28% for 200MB), but I think this should be done by default. The header+body is stored by storing UID 1=message 1 header, UID 2=message 1 body, UID 3=message 2 header, etc. I haven't yet tried how well the UID lists could be compressed if could support only-header or only-body ranges. For example now if word is found from all headers but from no bodies, it stores the UID list as 1,3,5,7,etc, while it could be stored as simple "only bodies, 1:7". I guess the UID list compression could use all kinds of other algorithms as well. Currently it supports only the whole list being either a bitmask or UID range. A long UID list might compress better as a bitmask +range combination. UID lists take most of the Squat space, so even small improvements can give huge benefits. If you're interested in trying, you can download the current test code and modify uidlist_write_array() for writing and node_uidlist_get_at_offset() for reading. http://dovecot.org/tmp/test.c http://dovecot.org/tmp/test-uidlist.c Place to src/plugins/fts-squat/, run with ./test some-mbox-file Another somewhat different idea to substring indexing (suggested to me by someone else) would be to index the messages' words normally like most indexers do, and then use Squat indexes for those words. So for example when searching for "ord" it first looks up all words that contain "ord" substring in them. Then it looks up message UIDs for those words, and the resulting UID list is a definite list of messages that has the substring. I'm not sure how well the above would work in practise. If the list of words would be mostly static it would work great. Unfortunately every mail has at least two unique strings: Message-ID and Received header's ESMTP ID. Dovecot must be able to search those as well. Also with some languages, like Finnish, a single word can be written in tens of different ways and combined with other words. A quick preliminary testing for 32MB Dovecot mailing list shows: - 142265 unique words takes 2,0MB, one per line - indexing those words creates 6,7MB Squat indexes So not including extra overhead for the unique word index, these use a total of 8,8MB. 4-char-substring Squat indexes for the same mailbox take 8,5MB and 4-char-substring+255-char-full-word Squat indexes take 11MB. So this doesn't look very promising.

1 0