[Dovecot] Message size incorrect?

Tue Jan 16 18:17:38 UTC 2007

On Tue, January 16, 2007 16:39, Timo Sirainen wrote:
> On Tue, 2007-01-16 at 11:00 +0100, Tristan Woudenberg - Prism Mail
> Solutions wrote:
>> There is quite a discrepancy in the rfc822.size reported by Dovecot and
>> the actual space it takes up on my harddisk. I know there could be some
>> overhead here from all the dovecot* and subscription list files in my
>> imap
>> folders, but like I stated in my email to the SquirrelMail plugin
>> mailing
>> list, about 1mb for a 4.3mb maildir mailbox might be a but much.
>
> RFC822.SIZE reports message's "virtual size" as it is sent to IMAP
> clients, which means that if the maildir files contain LF-only linefeeds
> (which is usually the case), the RFC822.SIZE increases the size by the
> number of lines in the file (so each line ends with CRLF).
>
> That however only increases the size reported by the Squirrelmail
> plugin. I don't know why it would report 1 MB less.
>
> How does Squirrelmail plugin's INBOX usage compare with "du -sh cur
> new"? Do you have any files in tmp/ directories? Currently Dovecot never
> cleans up the tmp/ directory. I guess I should add the code for that..

I already posted this on the Squirrelmail list, but the folder sizes
plugin outputs this:

Folder Sizes
Folder        Count        Unread        Size
INBOX        260        19        3.2 M
Drafts        0        0        0 k
Sent        0        0        0 k
Trash        0        0        0 k
spam        0        0        0 k
1 Folder        0        0        0 k
5 Folders        260        19        3.2 M

while when I go to that maildir folder on my server:

$ du -hs useraccount
4.3M   useraccount

Of course there's some overhead:

du -ah | grep dovecot
4.0K    ./.INBOX.spam/dovecot.index.log
4.0K    ./.INBOX.spam/dovecot.index
32K     ./.INBOX.spam/dovecot.index.cache
4.0K    ./.INBOX.spam/dovecot-uidlist
4.0K    ./.Drafts/dovecot.index.log
4.0K    ./.Drafts/dovecot.index
16K     ./.Drafts/dovecot.index.cache
4.0K    ./.Drafts/dovecot-uidlist
4.0K    ./.Sent/dovecot.index.log
4.0K    ./.Sent/dovecot.index
16K     ./.Sent/dovecot.index.cache
4.0K    ./.Sent/dovecot-uidlist
4.0K    ./.Trash/dovecot.index.log
4.0K    ./.Trash/dovecot.index
20K     ./.Trash/dovecot.index.cache
4.0K    ./.Trash/dovecot-uidlist
8.0K    ./dovecot.index.log
4.0K    ./dovecot.index
120K    ./dovecot.index.cache
8.0K    ./dovecot-uidlist

du -h subscriptions
4.0K    subscriptions

276K in total, 4.4M - 276K isn't the reported 3.2M I would guess. Also if
I do a FETCH 1:* (RFC822.SIZE) by hand on my imap server and sort the
results for the 10 biggest emails:

(RFC822.SIZE 57151)
(RFC822.SIZE 57352)
(RFC822.SIZE 57477)
(RFC822.SIZE 57498)
(RFC822.SIZE 57640)
(RFC822.SIZE 57776)
(RFC822.SIZE 58842)
(RFC822.SIZE 60778)
(RFC822.SIZE 61320)
(RFC822.SIZE 79539)

10 biggest files with du -ab

56526   ./cur/1165969700.640.mbox:2,
56754   ./cur/1166710093.P5305Q0M44492.host.domain.nl:2,S
56943   ./cur/1165969704.640.mbox:2,
57151   ./cur/1165969708.640.mbox:2,
57498   ./cur/1165969702.640.mbox:2,
57640   ./cur/1165969709.640.mbox:2,
57830   ./cur/1168348302.H160143P8248.host.domain.nl:2,
59710   ./cur/1166710169.P6010Q9M768683.host.domain.nl:2,S
61320   ./cur/1165969694.640.mbox:2,
79539   ./cur/1165969642.640.mbox:2,S

Quite strange don't you think? Some files do exactly match others don't, I
doublechecked this by grepping the output for one of these mail sizes.

Kind regards,

Tristan

-- 
Prism Mail Solutions
Deployment, customization and hosting
http://prism.nl