dovecot - Mar 2022 - Replicator Issues on Large Mail Boxes

Ok running :

freebsd-12.1
dovecot-2.3.18.current
dovecot-2.3-pigeonhole-0.5.18.current

simply put replication works fine on smaller email boxes without issues.

dsync also works when run manually, longest sync is 60 secords or so 
when using dsync so need replicator bumped up ?

i get the fact that the file locking issue relative since a larger 
mailbox will take longer to replicate when an email comes in ?

mail18????? 03-05 18:05:51 {dovecot}??????? [15799] (872623573) 
doveadm(keith at elirpa.com)<32249><GnS3M53sI2L5fQAAz1jc/w>: Error: 
Couldn't lock
/data/dovecot/users/elirpa.com/keith at elirpa.com//tmp/.dovecot-sync.lock:
fcntl(/data/dovecot/users/elirpa.com/keith at
elirpa.com//tmp/.dovecot-sync.lock,
write-lock, F_SETLKW)
 ??????????????????????????????????????????????????? locking failed: 
Timed out after 30 seconds (WRITE lock held by pid 31519)

so what i need is where fcntl is setting a 30 second timeout for 
replication (i have adjust all the others in the src code)

simply put replicator fails and retries and keeps failing which is 
understandable as it probably needs a little more time ?

# sync.users
carol at scom.ca???????????????????? none???? 00:02:51? 08:16:34 
-??????????? y
nick at elirpa.com?????????????????? low????? 02:45:17? 09:28:32 
-??????????? y
keith at elirpa.com????????????????? none???? 02:30:13? 09:28:32 
-??????????? y
paul at scom.ca????????????????????? high???? 02:45:17? 09:28:32 
-??????????? y
ed at scom.ca??????????????????????? none???? 02:34:34? 09:28:32 
-??????????? y
ed.hanna at dssmgmt.com????????????? high???? 02:45:17? 09:28:32 
-??????????? y

i found under /programs/src/mail/dovecot-2.3.18.current/src/lib

file-lock.c

 ??? struct flock fl;

 ??? fl.l_type = lock_type;
 ??? fl.l_whence = SEEK_SET;
 ??? fl.l_start = 0;
 ??? fl.l_len = 0;

 ??? ret = fcntl(fd, timeout_secs != 0 ? F_SETLKW : F_SETLK, &fl);
 ??? if (timeout_secs != 0) {
 ????? alarm(0);
 ????? file_lock_wait_end(path);
 ??? }

 ??? if (ret == 0)
 ????? break;

 ??? if (timeout_secs == 0 &&
 ??????? (errno == EACCES || errno == EAGAIN)) {
 ????? /* locked by another process */
 ????? *error_r = t_strdup_printf(
 ??????? "fcntl(%s, %s, F_SETLK) locking failed: %m "
 ??????? "(File is already locked)", path, lock_type_str);
 ????? return 0;
 ??? }

 ??? if (err_is_lock_timeout(started, timeout_secs)) {
 ????? errno = EAGAIN;
 ????? *error_r = t_strdup_printf(
 ??????? "fcntl(%s, %s, F_SETLKW) locking failed: "
 ??????? "Timed out after %u seconds%s",
 ??????? path, lock_type_str, timeout_secs,
 ??????? file_lock_find(fd, set->lock_method,
 ???????????????? lock_type));
 ????? return 0;
 ??? }
 ??? *error_r = t_strdup_printf("fcntl(%s, %s, %s) locking failed:
%m",
 ????? path, lock_type_str, timeout_secs == 0 ? "F_SETLK" :
"F_SETLKW");
 ??? if (errno == EDEADLK && !set->allow_deadlock) {
 ????? i_panic("%s%s", *error_r,
 ??????? file_lock_find(fd, set->lock_method,
 ???????????????? lock_type));
 ??? }
 ??? return -1;
#endif





-- 


Happy Saturday !!!
Thanks - paul

Paul Kudla


Scom.ca Internet Services <http://www.scom.ca>
004-1009 Byron Street South
Whitby, Ontario - Canada
L1N 4S3

Toronto 416.642.7266
Main?1.866.411.7266
Fax?1.888.892.7266
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<https://dovecot.org/pipermail/dovecot/attachments/20220305/73d06917/attachment-0001.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: scomca-logo.jpg
Type: image/jpeg
Size: 135491 bytes
Desc: not available
URL:
<https://dovecot.org/pipermail/dovecot/attachments/20220305/73d06917/attachment-0001.jpg>

Hi,

 > simply put replication works fine on smaller email boxes without issues.
 >
 > dsync also works when run manually, longest sync is 60 secords or so
 > when using dsync so need replicator bumped up ?

You can configure dsync parameters for the replicator to adjust the time 
limit, e.g.:


replication_dsync_parameters = -d -N -l 120 -U



best regards,
Carsten