AI Archive Directory error

Posted by James Palmer on 06-May-2015 05:57

Configuring up our new Win Server 2012 R2 server for Progress 11.5. Changing the way our AI Archiving works, to push the files onto a different server where they can be archived. Unfortunately, according to proserve, the location doesn't exist even if it actually does, and we have full control. 

                Wed May 06 11:54:52 2015
[2015/05/06@11:54:52.561+0100] P-3892       T-656   I BROKER  0: (333)   Multi-user session begin. 
[2015/05/06@11:54:52.562+0100] P-3892       T-656   I BROKER  0: (10545) Connections to this database will not be allowed until all Database Services started have completed their startup and initialisation. 
[2015/05/06@11:54:52.580+0100] P-3892       T-656   I BROKER  0: (5326)  Begin Physical Redo Phase at 0 . 
[2015/05/06@11:54:52.580+0100] P-3892       T-656   I BROKER  0: (7161)  Physical Redo Phase Completed at blk 0 off 235 upd 0. 
[2015/05/06@11:54:52.581+0100] P-3892       T-656   I BROKER  0: (13547) At end of Physical redo, transaction table size is 2048. 
[2015/05/06@11:54:52.633+0100] P-3892       T-656   I BROKER  0: (13187) The directory \\tera\progress\servers\hob\hobdbbackup\aiarchive\live specified with -aiarcdir does not exist. 
[2015/05/06@11:54:52.634+0100] P-3892       T-656   I BROKER  0: (13193) The after-image extent manager initialisation failed.  See previous errors for the reason. 
[2015/05/06@11:54:53.003+0100] P-3892       T-656   I BROKER   : (16869) Removed shared memory with segment_id: 40960000 
[2015/05/06@11:54:53.004+0100] P-3892       T-656   I BROKER   : (334)   Multi-user session end. 


I can create files etc manually in his location. 

Is there a limitation to the AI Archiver that the files have to be local? 

All Replies

Posted by Libor Laubacher on 06-May-2015 06:13

So there is no adminserver involved starting the db ?
 
[collapse]
From: James Palmer [mailto:bounce-jdpjamesp@community.progress.com]
Sent: Wednesday, May 6, 2015 12:59 PM
To: TU.OE.RDBMS@community.progress.com
Subject: [Technical Users - OE RDBMS] AI Archive Directory error
 
Thread created by James Palmer

Configuring up our new Win Server 2012 R2 server for Progress 11.5. Changing the way our AI Archiving works, to push the files onto a different server where they can be archived. Unfortunately, according to proserve, the location doesn't exist even if it actually does, and we have full control. 

                Wed May 06 11:54:52 2015
[2015/05/06@11:54:52.561+0100] P-3892       T-656   I BROKER  0: (333)   Multi-user session begin. 
[2015/05/06@11:54:52.562+0100] P-3892       T-656   I BROKER  0: (10545) Connections to this database will not be allowed until all Database Services started have completed their startup and initialisation. 
[2015/05/06@11:54:52.580+0100] P-3892       T-656   I BROKER  0: (5326)  Begin Physical Redo Phase at 0 . 
[2015/05/06@11:54:52.580+0100] P-3892       T-656   I BROKER  0: (7161)  Physical Redo Phase Completed at blk 0 off 235 upd 0. 
[2015/05/06@11:54:52.581+0100] P-3892       T-656   I BROKER  0: (13547) At end of Physical redo, transaction table size is 2048. 
[2015/05/06@11:54:52.633+0100] P-3892       T-656   I BROKER  0: (13187) The directory \\tera\progress\servers\hob\hobdbbackup\aiarchive\live specified with -aiarcdir does not exist. 
[2015/05/06@11:54:52.634+0100] P-3892       T-656   I BROKER  0: (13193) The after-image extent manager initialisation failed.  See previous errors for the reason. 
[2015/05/06@11:54:53.003+0100] P-3892       T-656   I BROKER   : (16869) Removed shared memory with segment_id: 40960000 
[2015/05/06@11:54:53.004+0100] P-3892       T-656   I BROKER   : (334)   Multi-user session end. 


I can create files etc manually in his location. 

Is there a limitation to the AI Archiver that the files have to be local? 

Stop receiving emails on this subject.

Flag this post as spam/abuse.

[/collapse]

Posted by James Palmer on 06-May-2015 06:30

Actually yes I'm using OEM (because that's the way we've always done it!).

Posted by Libor Laubacher on 06-May-2015 06:35

Then you need to change the account starting the adminserver (services – adminservice for 11.5 – properties – logon), by default is localsystemaccount which has no permissions to shares (and other stuff).
 
[collapse]
From: James Palmer [mailto:bounce-jdpjamesp@community.progress.com]
Sent: Wednesday, May 6, 2015 1:31 PM
To: TU.OE.RDBMS@community.progress.com
Subject: RE: [Technical Users - OE RDBMS] AI Archive Directory error
 
Reply by James Palmer

Actually yes I'm using OEM (because that's the way we've always done it!).

Stop receiving emails on this subject.

Flag this post as spam/abuse.

[/collapse]

Posted by James Palmer on 06-May-2015 06:37

Yes of course I do! Thanks Libor.

Posted by bremmeyr on 06-May-2015 08:26

I have had issues with the AI daemon accessing non-local drives even after providing a network account for adminserver. Following working with tech support the case was closed as an bug with a solution to be provided in a future version or service pack. Until then I have adjusted to archive to a local drive. Files are relocated to a remote drive without depending on the AI daemon.

OE10.2B Windows (various versions)

Posted by James Palmer on 06-May-2015 08:34

Thanks [mention:51801d8efe704cbdbfe4264ddf10d8f6:e9ed411860ed4f2ba0265705b8793d05] . Unfortunately due to the server configuration that is thrust upon me I don't have the storage available to me locally to archive locally. So far my testing seems to be going to plan though... :)

Posted by Paul Koufalis on 06-May-2015 08:36

I have seen databases FREEZE when the AI Archiver tried to write to a remote NFS directory. IIRC the AI Archiver grabbed some latch and never let go. The root cause was clearly identified as the NFS mount: it showed as available but you could not actually read or write to it. The enhancement request to Progress would be to more gracefully handle such a situation.

Since then, we never archive directly to a remote mount. We always archive locally then scp the file using a script.

Posted by James Palmer on 06-May-2015 08:37

Thanks Paul. I'll bare that in mind, and say "I told you so" when we encounter this issue. ;)

Posted by Rob Fitzpatrick on 06-May-2015 21:42

I've also hit the issue Paul talked about.  I couldn't disconnect the AIMD and had to bounce the database to resolve it.  Not much fun.  I now archive to a local directory and copy to DR.  

You don't need a lot of space for this James.  It isn't a permanent local archive.  You're just storing archived extents long enough to copy them off the box.

Posted by marcv on 07-May-2015 08:47

A similar problem can occur with online backups that write to remote storage. The first part of an online backup backs up the BI file and during this time backup process has a latch on the DB. If this takes too long it can wreak havoc with your application because all other DB access has to wait.

We just had an incident where for some reason the access to the NFS mounted backup area suddenly had delays. This caused the BI backup to go from seconds to minutes. When this happened all the app servers locked up after spawning their maximum configured agents.

This thread is closed