After 5 months, here is a new blog post :)

Yesterday, I was upgrading a customer server and I have noticed that DAOS folder was larger than it should be. Since my customer is heavily using attachments in messages, a large DAOS folder is expected but 150 GB attachments for less than 100 users? Something's wrong here.

Of course, DAOS wasn't in SYNCHRONIZED state which prevents PRUNE operations.

However, it's more important to find out why DAOS loosing synchronization?

At this point I noticed something very strange.

> tell daosmgr databases 
14.08.2013 03:04:34   DAOSMGR: Status DATABASES started
14.08.2013 03:04:34   DAOS database status:
...
...
14.08.2013 03:04:34   Database: S:\$RECYCLE.BIN\S-1-5-21-1801674531-1592454029-1177238915-500\$RD39NSB.nsf
14.08.2013 03:04:34   Database state = Synchronized
14.08.2013 03:04:34   Last resynchronized: 14.08.2013 02:40:02
14.08.2013 03:04:34   Ticket count: 239
...
...


There were lots of similar database records from the recycle bin which is supposed to be empty... So I have looked at that folder via command prompt and yes, those files really exist! After a research, I noticed the problem. Domino is looking for the NSF and NTF files outside the Data folder (even in recycle bin) to be handled by DAOS.

There are two common cases: First, most companies want to keep mail databases after an employee left. Since Domino doesn't provide any mechanism than deleting, they simply move the NSF file to a backup folder at the file level. Second, when needed, they restore a database into the DATA folder and delete them afterwards. In both ways, Domino (and DAOS) keep their track and continue to maintain these databases in DAOS catalog. Consequences?

1. File-level operations will change the status of DAOS to unsynchronized and stop PRUNE operations.
2. NLO files linked to those databases will never expire and you will see lots of unnecessary NLO files filling your precious storage space.
3. Those databases will stay open and locked by Domino processes so cannot be deleted. Worse, recycle bin would not be emptied like in our case: We found 2 GB dead space hidden inside the recycle bin.

Moreover, it's not limited to DAOS-enabled NSF files. "nserver" process will lock all nsf files in your system in case they can be used by DAOS.

Solution for the problem is being more organized for daily operations. Here are my suggestions:


Archiving mail databases for former employees:
  • If you keep those databases as DAOS-enabled, it doesn't make sense to archive them, because you will not be able to open attachments when needed.
  • We have to turn DAOS off and embed all attachments back in the database by "load compact mail\xyz.nsf -c -daos off"
  • Then we will change its extension from ".nsf" to ".nsfbackup" immediately after "dbcache flush" command.
  • Now we can remove and archive them as you wish.

Restoring a database to the server
  • Make sure you have read this wiki article... Read it again... Read carefully one last time. It's important :)
  • If you restore a very old database (older than your "Defer object deletion" value in the server config), NLO files linked to the database would be already pruned. So you should use documented methods to find necessary NLO files.
  • After restoring the database, be quick. Get your task done and delete the database from the data folder. When deleting, prefer using Administrator client. If you can't, change the file extension before deletion as above.

DAOS Catalog State
  • After file system operations, it's possible for DAOS to lose its syncronization state. You can check the state by "tell daosmgr status" command.
  • To resync, "tell daosmgr resync" command will work but remember, it's a CPU-intensive operation.
  • Use DDM to monitor DAOS state.

General Rules of Thumb
  • Never keep an NSF or NTF file outside of your DATA folder. Put your backup's into compessed files like ZIP or RAR or change their extensions.
  • Domino servers should not use file-level antivirus scanners or any other software that may lock files. If you have to use a real time scanner, use exclusions like explained here and here.
  • Always backup DAOS folder :)


Serdar Basegmez   |   August 14 2013 05:11:00 PM   |    DAOS  Problems  System Administration  Tips    |  
  |   Next   |   Previous

Comments (4)

Gravatar Image
Serdar Basegmez    http://lotusnotus.com    09/06/2013 11:47:11 PM

David,

I don't think you have the same issue. Domino-level deletion doesn't leave a trace as you said.

There should be another reason about losing RESYNC state. One case, backup software was such a case for me. Older versions of Domino have also some problems I know. Sometimes a single corrupted NSF might be a problem as well. Did you try fixup -J?

To find the root cause, there is a debug parameter.

DEBUG_DAOS_STATECHANGE=1

This might give an idea why it goes into NEED RESYNC status.

I hope it helps...

Gravatar Image
David Brown    http://port1352.com    09/06/2013 11:37:56 PM

I generally just create a program that runs nightly to try to combat the usually inexplicable DAOS desynchronization. I'm not using file-level AV and I approve adminP requests to delete mail databases that are no longer needed (so they aren't going to the recycle bin in my case). Still it is expected that DAOS will regularly lose sync. (Shame on IBM for not having this detected and resolved automatically.)

So, nserver.exe -c "tell daosmgr resync" solves the issue for me.

Are you suggesting that I am missing a root cause for it to lose sync in the first place?

Gravatar Image
Serdar Basegmez    http://lotusnotus.com    08/16/2013 9:47:08 PM

Paul, thx for your comment.

I disagree about it. It's all about optimization. I achieved real good savings using DAOS but it should be optimized carefully and maintained well.

If PRUNE operation fails, it will increase the size of the DAOS folder in time. About SPAM messages, you should keep minimum attachment size at a reasonable level. If it's too low, that will be a problem. Also, the defer deletion time is important. If you don't restore old mail databases frequently, you might configure that time as zero.

I have blogged about gateway mail servers before. You should not enable DAOS for mail.box if the mail server is acting as a gateway.

DAOS performs really well for especially e-mail servers with 50+ users...

Gravatar Image
Paul       08/15/2013 5:26:22 PM

DAOS has this and many other problems... it's IMPOSSIBLE to see if DAOS really saved space. At first you will see a nice savings but later during the next months you will notice that the DAOS folder will increase a lot. You cannot know why since there is no way to know to which DB a given NLO file belongs to.

Other problem with DAOS is that it also affect SPAM emails (SPAM emails contains images and images are internally attachments so they are caught by DAOS) so emails will be deleted as soon as they arrive but their storage will still be in the DAOS folder until next prune. So in practice I doubt that DAOS really helps to save a lot and it demans by far more administration efforts.