The code has some checks that if posix_fallocate() fails with a
specific errno it stops trying to use it. Maybe it hits that condition
at some point. Or maybe the code just isn't called for some reason, I
don't really know..
On Oct 9, 2008, at 6:45 PM, Stewart Dean wrote:
I have a call open to IBM with their Compiler group on this to see
if this can't be fixed right. A side question: how come is it that
this happens when the session starts up and reoccurs periodically
for the first day or so...and then not again unless and until those
imap process sessions are closed outTimo Sirainen wrote:
On Fri, 2008-10-03 at 14:33 -0400, Stewart Dean wrote:
I am seeing posix_fallocate and file_set_size errmsgs in the mail
syslog, but see a pattern:
- They only happen with the /var/spool/mail inbox NOT with any of
the /home folders and appear to be happening every 10 minutes from the time
I started DC (9AM, 10/1/98) until 11AM, 10/2...and then ceased The every ten minute message sets looked like this:Oct 1 22:30:31 egg mail:err|error dovecot: IMAP(sdean):
posix_fallocate() failed: Resource temporarily unavailableThe main problem here is that posix_fallocate() is broken in your AIX (v1.0 doesn't even try to use it). My previous patch attempted to
make Dovecot detect this and silently fallback to not using it, but apparently it can fail in more ways. I thought about adding another check for EAGAIN, but perhaps posix_fallocate() just returns the previous errno so it can't be checked that way. So I moved the
check to configure instead:http://hg.dovecot.org/dovecot-1.1/rev/12565ef10d1c
Alternatively you could just remove HAVE_POSIX_FALLOCATE from
config.h after running configure. Or yet another way would be to try to find
out if it's already been fixed in AIX. This looks related: http://www-01.ibm.com/support/docview.wss?uid=isg1IY77112
- However, then there was the following: a) If I used webmail, which accessed the production server and got
the indices on my test server out of sync, I got this error message from in
the mail syslog on my test server:Oct 3 12:20:23 egg mail:err|error dovecot: IMAP(sdean): mbox
sync: UID inserted in the middle of mailbox /var/spool/mail/sdean (648818 > 648046,
seq=1153, idx_ msgs=1187)v1.1 also has a bug that can cause this, although normally it
should be visible only when index files aren't being used, or they're out of
sync for some reason. This'll fix it: http://hg.dovecot.org/dovecot-1.1/rev/a5bf7e12f3ccOct 3 12:44:58 egg mail:info dovecot: imap-login: Maximum number
of connections from user+IP exceeded: user=<sdean>, method=PLAIN,
rip=10.20.10.169, lip=192.24 6.229.31Turns out I had 10+ sessions, one back from yesterday, so I killed
them all and could get mail, but...about six minutes later, I had the two
posix_fallocate and file_set_size errmsgs again after not having any for a day. So
something about new connections maybe causes this?Any ideas why: a) I am having leftover IMAP sessions on my test server? This
doesn't happen on my production DC V1.0 serverAre you sure? Perhaps you just didn't notice them since v1.0 didn't
have any limits to how many were allowed? I think it's more likely that
the client(s) really just left that many connections. So the choices are:a) Increase mail_max_userip_connections setting.
b) Figure out where the sessions are from and see if you can do something about them on the client side. In Thunderbird there's a setting which specifies how many connections it can use.