Quotas for Postfix policyd
How to configure quotas for Postfix policyd (sqlite).
PolicyD v2 (codenamed "cluebringer") is a multi-platform policy server for popular MTAs. The main goal is to implement as many spam combating and email compliance features as possible. The configuration of policyd under Zimbra is done by editing the sqlite database as shown in the examples below.
NOTE: Currently the only supported way of configuring policyd is through the CLI. The policyd gui that comes with the package is not supported.
To start with the topic, we must first enable policyd to be able to use it:
zmprov ms <mta server> +zimbraServiceEnabled cbpolicyd
As we mentioned this article explains configuring policyd with sqlite. This database contains couple of tables, which are used for the configuration of policyd:
sqlite> .tables access_control greylisting_tracking accounting greylisting_whitelist accounting_tracking policies checkhelo policy_group_members checkhelo_blacklist policy_groups checkhelo_tracking policy_members checkhelo_whitelist quotas checkspf quotas_limits greylisting quotas_tracking greylisting_autoblacklist session_tracking greylisting_autowhitelist
We will predominantly work with the policies, policy_members, policy_groups, policy_group_members, quotas and quotas_limits tables. First, we need to navigate to and enter the sqlite database:
$ cd /opt/zimbra/data/cbpolicyd/db/ $ sqlite3 cbpolicyd.sqlitedb SQLite version 3.6.20 Enter ".help" for instructions Enter SQL statements terminated with a ";" sqlite>
To configure policyd, it's important to understand the logic flow and the schema structure of the tables, as it might be confusing at the beginning. With the examples below I will also include the schema structure to understand the flow better. There are two possibilities: to configure policy with groups or without groups.
For both possibilities we need to do these two steps first:
1. Create a policy in the policies tables. 2. Configure the policy members of the newly created policy in the policy_members table.
a) If we are not to use groups, then after configuring the policy and policy members, we edit the quota and quotas_limits tables. b) If you are to configure groups, then after configuring the policies, we configure the groups and then the quotas.
- Creating a policy
The following command creates policy with name "test_policy", with priority 0 and is enabled:
sqlite> insert into policies(Name,Priority,Disabled) VALUES ('test_policy',0,0);
Priority 0, means it will be picked up first from the list of the policies. The priorities goes as 1, then 2 and so on.
Example of our newly created policy (the ID,Name etc are not included in the normal view, but I put them to see the column name and corresponding data):
sqlite> select * from policies; ID|Name |Priority|Description |Disabled 1|test_policy|0 | |0
- Creating policy members
The next thing to do is to edit the policy_members table, to specify the members that will be included in the policy. We will give two examples: one using group and without.
sqlite> insert into policy_members(PolicyID,Source,Destination,Disabled) VALUES (1,'%test_group','any',0);
sqlite> insert into policy_members(PolicyID,Source,Destination,Disabled) VALUES (1,'firstname.lastname@example.org','any',0);
The result from these two commands is:
sqlite> select * from policy_members; ID|PolicyID|Source |Destination |Comment|Disabled 1|1 |%test_group |any | |0 2|1 |email@example.com|any | |0
Important to note here is the second column of the policy_members table. It points to the ID number of the policies table. In this case, both entries in the policies_members table point to the same policy: 1, from the policies table.
Another important thing to note is that you can not add two groups in the Source column for one ID. For example:
sqlite> select * from policy_members; ID|PolicyID|Source |Destination |Comment|Disabled 1|1 |%test_group1,%test_group2|any | |0
In the above example we see two groups: test_group1 and test_group2. In that case, the quota set for policy_member ID 1 will not enforce its rules. To configure two groups to point to the same PolicyID, we need to insert another entry into the policy_members table:
sqlite> select * from policy_members; ID|PolicyID|Source |Destination |Comment|Disabled 1|1 |%test_group1|any | |0 2|1 |%test_group2|any | |0
Here are the possible options that can be in the Source and Destination columns:
NULL = any a.b.c.d/e = IP address with optional /e @domain = domain specification, %xyz = xyz group, abc@domain = abc user specification all options support negation using !<key>
To get these values for any of the tables, you can use the .schema command:
sqlite> .schema policy_members CREATE TABLE policy_members ( ID INTEGER PRIMARY KEY AUTOINCREMENT, PolicyID INT8, /* Format of key: NULL = any a.b.c.d/e = IP address with optional /e @domain = domain specification, %xyz = xyz group, abc@domain = abc user specification all options support negation using !<key> */ Source TEXT, Destination TEXT, Comment VARCHAR(1024), Disabled SMALLINT NOT NULL DEFAULT '0', FOREIGN KEY (PolicyID) REFERENCES policies(ID) );
- Adding group/s
1. Adding group is similar to adding policies, just the table columns' names are different. In our previous example, we have entered a group (test_group) in the policy_members table. However the group is not created yet. Here is how to do it:
sqlite> insert into policy_groups(Name,Disabled) VALUES ('test_group',0);
Here is the group in sqlite:
sqlite> select * from policy_groups; ID|Name |Disabled|Comment 1|test_group|0 |
2. The next step is to create the members for the test_group. Below is a part of the .schema command for this table, which shows us the correct syntax for the policy_group_members table:
/* Format of member: a.b.c.d/e = ip, @domain = domain, %xyz = xyz group, abc@domain = abc user */
To add couple of members, we can run the following commands:
sqlite> insert into policy_group_members(PolicyGroupID,Member,Disabled) VALUES (1,'testuser@mydomain',0); sqlite> insert into policy_group_members(PolicyGroupID,Member,Disabled) VALUES (1,'192.168.56.10/24',0);
And the result is:
sqlite> select * from policy_group_members; ID|PolicyGroupID |Member |Disabled|Comment 1|1 |testuser@mydomain|0 | 2|1 |192.168.56.10/24 |0 |
The above is showing that all quotas, pointing to policy ID 1, will be forcing their rules on testuser@mydomain, machine with ip 192.168.56.10 and firstname.lastname@example.org, specified as ID 2 in the policy_members table. Also as in the previous example, the PolicyGroupID is pointing to a specific policy_groups entry ID. In our case it's pointing to ID 1.
- Configure quotas
To configure quotas, we will have to edit two tables: quotas and quotas_limits. First we will edit the quotas table, but before that we need to check the formatting:
/* Tracking Options */ /* Format: <type>:<spec> SenderIP - This takes a bitmask to mask the IP with. A good default is /24 Sender & Recipient - Either "user@domain" (default), "user@" or "@domain" for the entire email addy or email addy domain respectively. */
And the command to create a quota:
sqlite> insert into quotas(PolicyID,Name,Track,Period,Verdict,Data,Disabled) VALUES (1,'test_quota','Sender:user@domain',120,'DEFER','User has been deferred for two minutes',0);
Followed by command to edit the quota_limits table:
sqlite> insert into quotas_limits(QuotasID,Type,CounterLimit,Disabled) VALUES (1,'MessageCount',2,0);
The above commands are enforcing quota that is restricting all users in %test_group and email@example.com to send no more than two messages in a time frame of two minutes. Notice that the quota is pointing to a specific policy. We have only one policy, so its pointing to policy ID 1. The quota_limits table has an entry that is pointing to the quota from the quota table. Instead of "Sender", we can specify "Recipient", which will limit the emails received by user/group.
- Below is a little example of already created policy/group/quota:
sqlite> select * from policies; ID|Name |Priority|Description|Disabled sqlite> 1|test_policy|0 | |0 - sqlite> select * from policy_members; ID|PolicyID|Source |Destination|Comment|Disabled sqlite> 1|1 |%test_group |any | |0 sqlite> 2|1 |firstname.lastname@example.org|any | |0 - sqlite> select * from policy_groups; ID|Name |Disabled|Comment sqlite> 1|test_group|0 | - sqlite> select * from policy_group_members; ID|PolicyGroupID|Member |Disabled|Comment sqlite> 1|1 |%group2 |0 | sqlite> 2|1 |192.168.56.10/24|0 | - sqlite> select * from quotas; ID|PolicyID|Name |Track |Period|Verdict|Data |LastQuota|Comment|Disabled sqlite> 1|1 |test_quota|Sender:user@domain|120 |DEFER |User has been deferred for two minutes|0 | |0 - sqlite> select * from quotas_limits; ID|QuotasID|Type |CounterLimit|Comment|Disabled sqlite> 1|1 |MessageCount|2 | |0
Let's try to explain the above configuration:
-1. We have one policy: test_policy. -2. In the policy_members table, we have two members pointing to the test_policy. (notice the PolicyID number being the same as the ID of the policy). In the same table we have specified one group (indicated by the % sign). This group is created in the policy_groups table. -4. policy_group_members table contains the ips, domains, groups to which to apply the quotas. -5. The quotas table specify the name of the quota, to which property (track) to apply the quotas, the time period for which the quota is applied, Verdict as to what to do if the quota has a match and if its disabled or not. -6. The quota_limits, specifies the messages limits that could be sent/received.
There are couple of other options that can be configured. They can be seen if you check the schema for the table you would like to modify. For example:
sqlite> .schema quotas_limits CREATE TABLE quotas_limits ( ID INTEGER PRIMARY KEY AUTOINCREMENT, QuotasID INT8, Type VARCHAR(255), /* "MessageCount" or "MessageCumulativeSize" */ CounterLimit UNSIGNED BIG INT, Comment VARCHAR(1024), Disabled SMALLINT NOT NULL DEFAULT '0', FOREIGN KEY (QuotasID) REFERENCES quotas(ID) );
As we can see, for the quota_limits we can also specify "MessageCumulativeSize", which means "Cumulative size of email messages, or you could say bandwidth quota. This is in bytes."
And finally, if you prefer to disable the service, you can run:
zmprov ms <mta server> -zimbraServiceEnabled cbpolicyd