Dear piler users,

I have a single idea in mind before release the final 0.1.25 version of piler. The gui supports groups, so you can make arbitrary permissions for users. However currently the piler groups support local users only. I bet that most of you keep users in Active Directory, ldap or other external databases, so it would be
better to assign groups to them directly.

One possible method to do so is to redesign the group feature with the following
tables:

1. group: define the groups, eg. id=1, name=somenameforyourgroup
2. group_email: define emails assigned to the given group

Currently we have group_user to define local userid vs. group id associations,
eg.

MariaDB [piler]> select * from `group_user`;
+----+-----+
| id | uid |
+----+-----+
|  1 |   2 |
+----+-----+
1 row in set (0.00 sec)


Now I'd change this to replace the uid field with an email address. Then when an AD user logs in the gui queries all his email addresses, then we need another query
to determine any additional email address that the user needs access.

My question is whether you can accept such a change for your existing archives.

I also tried to improve the "Archive size:" on the health page. This number comes from the counter table, and it's the sum of the sizes of the received emails (before storing them, ie. without compression and deduplication). To make things clearer
my test archive has the following:

Archive size:   128.2 GB (59.0 GB)

So the archive stores 128.2 GBs of emails on 59 GB disk space only. That's why you see a different number at the disk usage. The "59 GB" also comes from the counter table, it has a new column which updates the stored size. One small issue for existing archives: to get a rather accurate result you should set some value for it by
estimating your current usage, eg.

update counter set stored_size = 59000000000;

But first extend the counter table:

alter table `counter` add column `stored_size` bigint unsigned default 0;


I've also updated the download page with rc2. rc1 has an unpleasant surprise. The duplicate detection uses a temp file. The trouble is that I forgot to add the matching close() syscall, so you may run out of the available open files limit, especially with using pilerimport. The
rc2 version fixes this issue.

So the plan is to give you 2 weeks to test rc2, and if you don't report issues, then I'll
release the final version.

If you have any questions, just let me know.

Janos

Reply via email to