Re: Loopback filesystems for mail storage
On Sat, Aug 17, 2013 at 02:28:46PM -0500, Stan Hoeppner wrote:
> And of course the simple solution is to migrate to mbox. Maildir was
> created for a single purpose: to eliminate file locking contention on
> the traditional UNIX single mbox file. With tens of thousands of emails
> this is obviously an archival workload. As such you don't have the
> locking issues. Header and body searches are many times faster with
> mbox than with maildir. The only operation that is significantly slower
> is deleting an email. I'll assume mutt uses mmap to access mbox files,
> so large file sizes shouldn't be a problem.
Thank you for your response, Stan.
The reason why I moved to Maildir is because I started using notmuch
for indexing and searching my e-mail. Now, I could maintain two
copies: one for general access and the other for notmuch indexing, but
that's rather inefficient and I'd have to sync them somehow.
> FWIW, I have some 100K+ emails, mostly list mail, in a couple dozen
> active mbox files. Full body searching is relatively quick on a file
> containing 20K emails and on a gzipped file containing 17K+ emails, even
> with stale FTS indexes in both cases. These mbox mails are stored on a
> Dovecot server w/dual 550MHz CPUs, 384MB RAM, with a single effective
> 7.2K SATA spindles--pretty old and slow hardware. mbox performance is
> more than acceptable.
Agreed. I'll see if there is a way by which I can have the best of
both worlds.
My usual solution (even with mboxes) was to use formail to split the
messages to several mail directories on a per month basis, such as
2013_08, 2013_07 … 2005_06 etc. I'll see if that helps.
Thanks.
Kumar
--
martin@bdsi.com (no longer valid - where are you now, Martin?)
-- from /usr/src/linux/drivers/cdrom/mcd.c
Reply to: