Gromox http timeouts / too much connections

ozy

Got the following in my logs:

oct. 14 15:29:32 email gromox-midb[24355]: E-2403: mail_engine: timed out obtaining a reference on /var/lib/gromox/user/domain.tld/user
oct. 14 15:29:34 email gromox-http[1680]: Rejecting connection from [::ffff:127.0.0.1]:52304: reached 1000 connections (http.cfg:context_num)
oct. 14 15:29:43 email gromox-http[1680]: Rejecting connection from [::1]:56606: reached 1000 connections (http.cfg:context_num)
oct. 14 15:29:45 email gromox-http[1680]: Rejecting connection from [::ffff:127.0.0.1]:43376: reached 1000 connections (http.cfg:context_num)

I've indeed set http.cfg:context_num to 1000 earlier.
This happened while I tried to soft-delete mails for a user that got like 120K mails in a couple of hours because of a config error.

Are there any best practices for small environments (couple of users) with 500K+ emails on IMAP ?

crpb

ozy Are there any best practices for small environments

I don't have any.
But if we get a couple of things together here that makes sense for the majority i'm sure we could make them sticky.

ozy

I still get errors like the following on some of my servers

oct. 13 11:34:42 email gromox-midb[1693]: E-2403: mail_engine: timed out obtaining a reference on

I noticed that this happens of two fairly big mailboxes (automatic emails sent X times a day, mboxes with 250K emails).
Are there any processes which could lock the midb sqlite tables ?

crpb

ozy Are there any processes which could lock the midb sqlite tables ?

lsfd -Q '(USER == "gromox") and (NAME =~ "^/var/lib/gromox.*sqlite3")'

you might leave out the USER == of course but haven't seen anything else

i only ever see http for exchange.sqlite3 and midb for midb.sqlite3

ozy

Nice, didn't know lsfd. I used lsof | grep midb earlier.
Got the same results, only gromox-midb holding those sqlite files.