Hi,
Thanks your reply, it moves me a bit forward.
Probably my problem was solved in:
commit ef0c4f10e897ec766e826b08cb6d75e49b68615a
Author: Ken Murchison <mu...@andrew.cmu.edu>
Date: Fri Oct 16 08:32:20 2015 -0400
ctl_mboxlist.c: properly dump/undump location field
(server!partition) for proxy servers
so it's on master since Oct. 2015.
I only checked the tarball release date for 2.5.10, which is Aug 31 2018:
(https://github.com/cyrusimap/cyrus-imapd/releases/tag/cyrus-imapd-2.5.10,
elliefm <https://github.com/elliefm>released thison Aug 31 2018)
So I assumed that the modification from 2015 must be in it, but it isn't.
I've downloaded the latest 2.5.17 tarball from here:
https://github.com/cyrusimap/cyrus-imapd/releases/tag/cyrus-imapd-2.5.17
elliefm <https://github.com/elliefm>released this21 days ago
and this still doesn't contains the desired modification from Ken
Murchinson:
In the tarball:
case DUMP:
if (!d->partition || !strcmpsafe(d->partition, mbentry->partition)) {
printf("%s\t%d %s %s\n", mbentry->name, mbentry->mbtype,
mbentry->partition, mbentry->acl);
if (d->purge) {
mboxlist_delete(mbentry->name, 0);
}
}
My desired patch:
commit ef0c4f10e897ec766e826b08cb6d75e49b68615a
Author: Ken Murchison <mu...@andrew.cmu.edu>
Date: Fri Oct 16 08:32:20 2015 -0400
ctl_mboxlist.c: properly dump/undump location field
(server!partition) for proxy servers
diff --git a/imap/ctl_mboxlist.c b/imap/ctl_mboxlist.c
index 4ac3b95bb..f6e9f5a43 100644
--- a/imap/ctl_mboxlist.c
+++ b/imap/ctl_mboxlist.c
@@ -178,8 +178,9 @@ static int dump_cb(const mbentry_t *mbentry, void
*rockp)
switch (d->op) {
case DUMP:
if (!d->partition || !strcmpsafe(d->partition,
mbentry->partition)) {
- printf("%s\t%d %s %s\n", mbentry->name, mbentry->mbtype,
- mbentry->partition, mbentry->acl);
+ printf("%s\t%d ", mbentry->name, mbentry->mbtype);
+ if (mbentry->server) printf("%s!", mbentry->server);
+ printf("%s %s\n", mbentry->partition, mbentry->acl);
if (d->purge) {
mboxlist_delete(mbentry->name, 0);
}
This is a bit confusing.
Why neither 2.5.10 or 2.5.17 tarball don't contains the mentioned patch?
I've built 2.5.17 from tarball, noticed the faulty behavior, obviously:
imre@main:~/work/cyrus-imapd-2.5.17$ sudo imap/ctl_mboxlist -d | grep
imre@ill
illusion.hu!user.imre 1 common i...@illusion.hu lrswipkxtecda
The git master works properly:
imre@main:~/work/cyrus-imapd$ sudo imap/ctl_mboxlist -d |grep imre@ill
illusion.hu!user.imre 1 cyrus-common!common i...@illusion.hu
lrswipkxtecda
I've also tried your PR 3384, which dumps a bit more information:
imre@main:~/work/cyrus-imapd$ imap/ctl_mboxlist -f
/home/imre/mailboxes.db -d | grep imre@ill
illusion.hu!user.imre 1 cyrus-common!common i...@illusion.hu
lrswipkxtecda >(null) 0 0 0 0 (null)
Is it safe to use the ctl_mboxlist built from master on a 2.5.10
Ubuntu built in prod environment?
If not, I assume that I have to build my own Ubuntu package with only
that one patch.
I'm just wondering how many bugfixes still missing in pre-build packages.
i.
2021.03.30. 12:33 keltezéssel, Дилян Палаузов írta:
Hello,
you do not want to do this, since ctl_mboxlist does not export all
fields in the database. So if you export, modify and import, some
fields are gone. Moreover I do not think this shall be modified on
the murder server. Rather the instances shall put changes to the
murder server, when it comes to the mboxlist.
On this occasion you might want to apply
https://github.com/cyrusimap/cyrus-imapd/pull/3384
<https://github.com/cyrusimap/cyrus-imapd/pull/3384> to modify your
ctl_mboxlist and see if the patch makes problems.
I cannot tell you why is ->server empty. You can try with
cvt_cyrusdb and convert to flat. It is not specific to the mailbox
database and extracts whatever is in the store.
Version 2.5.10 was released in 2016. I recommend you to see if more
recent versions can handle your case.
Greetings
Дилян
On Tue, 2021-03-30 at 02:05 -0400, i...@illusion.hu wrote:
Hi,
I have a discrete cyrus setup with two frontends, several backends
and a single murder.
On the murder server I would like to dump out the mailboxes.db into
a plain file, for doing sed/grep modifications then push it back.
I'm facing with an inconsistence where the dump made on murder does
not contains the backend's hostname.
on murder :
ctl_mboxlist -d |grep 'my own mailbox'
illusion.hu!user.imre 1 common i...@illusion.hu
<mailto:i...@illusion.hu> lrswipkxtecda
It tells me that is a remote (1) mailbox on partition common. But
on which backend host?
on backend:
ctl_mboxlist -d
illusion.hu!user.imre 0 common i...@illusion.hu
<mailto:i...@illusion.hu> lrswipkxtecda
looks good, it's a local mailbox on common partition
on frontend:
ctl_mboxlist -d
illusion.hu!user.imre 1 cyrus-common!common i...@illusion.hu
<mailto:i...@illusion.hu> lrswipkxtecda
looks good, it's remote (1) on host 'cyrus-common' on partition
'common'
On murder the mailboxes.db does contain the missing information. I
converted to db file to flat, got this:
illusion.hu!user.imre 1 cyrus-common!common i...@illusion.hu
<mailto:i...@illusion.hu><FF><89>lrswipkxtecda<FF><89>
On murder I'm using 2.5.10-3ubuntu1.1 version (ubuntu Bionic build)
I digged into the code, imap/ctl_mboxlist.c +183:
case DUMP:
if (!d->partition || !strcmpsafe(d->partition,
mbentry->partition)) {
printf("%s\t%d ", mbentry->name, mbentry->mbtype);
if (mbentry->server) printf("%s!", mbentry->server);
printf("%s %s\n", mbentry->partition, mbentry->acl);
if (d->purge) {
mboxlist_delete(mbentry->name);
}
}
how can the mbentry->server be empty in my case?
The site is operational, it works fine in live, but after a
dump/undump in murder all my clients started to complain about
dissappeared mailboxes. That made evidence that the frontend
configurated properly, using the murder's database as a master.
This problem is serious for us:
- one of our backend has destroyed, we need to purge the unused
mailboxes from database
- we want to change the hostname of a backend and want to avoid
mailbox deletion after 'mupdate -m'
- our mailboxes.db backup on murder is useless.
Please give me advices how to handle this case. If any detailed
information is needed I'm happy to send.
