Link to home
Start Free TrialLog in
Avatar of CMES-IT
CMES-ITFlag for United States of America

asked on

Named Property Limits on Exchange 2007 SP1

I have recently started receiving the EventID 9667, msgidNamedPropsQuotaError on a Exchange 2007 mailbox server.  Research has shown that the server has reached the default quotas for named propertiy entries for the database.  These can be increased by adjusting the values in the registry of the "Named Props Quota" and "NonMAPI Named Props Quota" for a mailbox database or "Replids Quota" for a public database.

KB articles say the default limits are 16384 for the "Named Props Quota"(NPQ) and "Replids Quota"(RQ) and 8192 for the "NonMAPI Named Props Quota"(NNPQ).  They also say that each database has a hard limit of 32766.  They say if you reach that limit, you will have to create new databases and move the users out and then back in to reset the numbers.  They also say you should not set the quotas to the max limits.  However, the instructions on how to change the values say to enter a number between 1 and 32767.

Currently, my registry only has the NNPQ setting with a value of 8192.  The 8192 value is the one being referenced in my event log errors.

Relevant articles:
Event ID 9667:  http://technet.microsoft.com/en-us/library/bb851495.aspx
How to configure:  http://technet.microsoft.com/en-us/library/bb851493.aspx
What they mean:  http://technet.microsoft.com/en-us/library/bb851492.aspx

What they don't say is how the numbers relate to each other.

Since the NPQ and NNPQ both affect the mailbox database, do their combined values need to stay below 32766 or are their limits separate?  
Do I only need to update the value that is already in the registry or should I be creating the others?

Thanks.

Avatar of Jack_French
Jack_French

More than likely you are reaching the max on the Non Mapi named props, I think you are saying that you believe that as well. When you send emails that cross spam filters or whatever, non mapi named props are added such as x-headers and so on, anyways, the two types are seperate, so yes you can just raise the single entry in the registry for non mapi to buy you more time. However....like you mentioned you should NOT set this to the MAX, because the next time you get the warning there will be no more time to add more grow room. I would recommend that you push the non mapi to 16k or so, then start your plans on how you will begin moving mailboxes to a new db.

How long have you had this DB in production? That is how long it took the DB to reach 8k, so that can help you determine how long you have before 16k....though this can spike and move faster and slower depending on mail traffic.

Also, the named prop tables are db specific so you will have to move the users to a new db, and you can not move them back into the old.

hope this makes sense, let me know if you have any questions.

jack
Avatar of CMES-IT

ASKER

If I bump the single for nonMAPI to 16k and the MAPI is already at 16k, won't that risk me reaching the set limit of 32k or are the limits seperate so they can both go to 32k?  I guess I'm asking if it is a combined limit or a seperate limit for each type?

They say that if you reach the 32k limit you should create a new db, move the users, delete and recreate the old db (makes a new file) and then move them back.

Thanks.

ASKER CERTIFIED SOLUTION
Avatar of DraconianSoul
DraconianSoul
Flag of United States of America image

Link to home
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
Start Free Trial
Avatar of CMES-IT

ASKER

That would make sense.  It would also explain why there is no value for the MAPI in the registry since it is already at the max by default.

As a side note:  From what I can see, the numerous headers being placed by spam control efforts combined with the spoofed headers in spam and virus messages, is causing this number to rise quicker than developers expected.  I have also heard rumor that rollup 7 may have a method to address hitting the limits that is not as drastic as the method currently required.  I guess we can wait and see.

Thanks,