How about you try moving the mail into another folder on daily basis,
this way the INBOX would stay nice and empty.
doveadm move -u arch...@company.com Archive MAILBOX INBOX SENTBEFORE
todays-date
Aki
On 16.02.2018 06:19, Terence Lau wrote:
>
> Bump.
>
>
>
> Any advice would be most appreciated.
>
>
>
> Thanks.
>
>
>
> *From:*Terence Lau
> *Sent:* Wednesday, 24 January 2018 9:59 AM
> *To:* 'dovecot@dovecot.org' <dovecot@dovecot.org>
> *Subject:* Out of memory on lmtp vsz_limit
>
>
>
> Hi,
>
>
>
> We’ve been getting these types or errors for quite a while now …
>
>
>
> Fatal: master: service(lmtp): child 63477 returned error 83 (Out of
> memory (service lmtp { vsz_limit=4096 MB }, you may need to increase it)
>
>
>
> … and these errors have been decreasing in occurrence as we increased
> the default_vsz_limit. Which is good but I would like to get some
> advice on how I could possibly eliminate the errors.
>
>
>
> We have an internal smtp server (postfix 3.1.0) that has the config
> “always_bcc=arch...@company.com
> <mailto:always_bcc=arch...@company.com>” over lmtp. This mailbox is
> on a separate dovecot server with the following config (please let me
> know if the full config is required):
>
>
>
> # 2.2.22 (fe789d2): /etc/dovecot/dovecot.conf
>
> # Pigeonhole version 0.4.13 (7b14904)
>
> # OS: Linux 4.4.0-109-generic x86_64 Ubuntu 16.04.1 LTS ext4
>
> default_vsz_limit = 4 G
>
> mail_location = maildir:/home/vmail/%d/%n
>
> protocols = " imap lmtp pop3"
>
> service lmtp {
>
> inet_listener lmtp {
>
> port = 24
>
> }
>
> }
>
> userdb {
>
> args = username_format=%u /etc/dovecot/users
>
> default_fields = uid=vmail gid=vmail home=/home/vmail/%d/%n
>
> driver = passwd-file
>
> }
>
> protocol lmtp {
>
> mail_plugins =
>
> }
>
>
>
> Since we discovered the errors, we’ve been increasing the
> default_vsz_limit to 1G, then 2G and now 4G (Server has 6GB of
> memory). These errors occur whenever a large number of emails get
> sent around the same time to our smtp server. This causes the dovecot
> server to start crunching CPU and Memory. Load average goes through
> the roof and takes some time to come back down as the smtp queue
> clears itself.
>
>
>
> This mailbox is obviously very large but we have a script that runs
> daily to delete any emails older than a month:
>
>
>
> find /home/vmail/company.com/archive/new/ -type f -mtime +30 -exec rm
> {} \;
>
> find /home/vmail/company.com/archive/cur/ -type f -mtime +30 -exec rm
> {} \;
>
>
>
> Still, the mailbox has on average of just under 300,000 emails. No
> one accesses this mailbox with an email client, not until we need to
> dig something up. And this has only happen once. So the emails
> pretty much never get read/process by a user.
>
>
>
> Now that we’ve increased the default_vsz_limit to 4G, the occurrence
> of these errors has reduced considerably. But they still happen
> occasionally. Short of increasing the memory further, are there any
> other options I have?
>
>
>
> Thanks.
>
